Re: [R-sig-phylo] data anlysis
Dear Rodrigo, Four species is too few to be able to tell with any confidence whether a statistical model fitted with a star phylogeny or a hierarchical phylogeny better fits your data. However, you can certainly analyze your data with both conventional and phylogenetic methods (e.g., independent contrasts, simulations). If the main result is similar with both approaches, then you have your answer. If not, then ... With that small a sample size, I would not bother with approaches that transform the branch lengths. Here are two papers that present simulations related to sample size: Blomberg, S. P., Garland, Jr., T. and Ives, A. R. (2003). Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57, 717–745. Freckleton, R. P., Harvey, P. H. and Pagel, M. (2002). Phylogenetic analysis and comparative data: a test and review of evidence. The American Naturalist 160, 712–726. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Rodrigo Gavira [rodgav...@gmail.com] Sent: Sunday, July 10, 2016 12:38 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] data anlysis Dear guys, my name is Rodrigo, I am a PhD student from São Paulo State University, Brazil. I am finishing my thesis but have some doubts about how to analyze my data, so I decided contact you. Here I am working on comparative physiology of ectothermic vertebrates. In summary, I work with 4 pitvipers species (*Bothrops* and *Crotalus* genus) in which I measured: Metabolism, water loss by evaporation, and thermal preference and tolerance. Now I would like to compare these results among species. However, as I said, I have only four species to compare. From what I understand when looking at mailing linst archives, this number of taxa are not enough to test the effects of phylogeny on my results, and I must use conventional analyzes. Is that right? If so, please, could you send me some papers dealing with this? Many thanks, guys. Cheers, -- MSc. Rodrigo Samuel Bueno Gavira Biologist - PhD student in Zoology Universidade Estadual Paulista - UNESP Department of Zoology Rio Claro - SP - BRAZIL [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] simulating continuous data
If you are trying to mimic real data, then perhaps you have some fossil data to go on? If not, then you can try to pick a "reasonable" value based on other biological knowledge. Check the Garland et al. (1993) for how we did it. Cheers, Ted From: Bryan McLean [bryansmcl...@gmail.com] Sent: Tuesday, May 10, 2016 3:32 PM To: Theodore Garland Jr Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] simulating continuous data Thanks Joe and Ted, By similar scaling, I just meant (as Ted guessed) that the root value depends on the empirical trait data, and does not start at 0 or 1, e.g., and thus produces simulated values that can be directly compared to the true data. Under a Brownian model, the mean trait value is suitable as the root value, but how does one specify a root value under a different and potentially better fitting model (OU, EB)? Im working mostly in R. -Bryan On 10 May2016, at 4:38 PM, Theodore Garland Jr > wrote: This is a good point and one that is often glossed over. We talked about it quite a bit here: Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265�292. http://www.biology.ucr.edu/people/faculty/Garland/GarlEA93.pdf Surely you want to do various descriptive statistics on your simulated data sets to see how they compare with the real one, and presumably you want some of those to include the phylogenetic versions (e.g., conventional and phylogenetic estimates of the correlation coefficient if you are simulating two traits). I think it is also really important to consider models that have limits to trait evolution (again, see the paper listed above). Those limits can interact strongly with starting (root) values, especially if you include evolutionary trends. Check Figure 1 in this paper: Diaz-Uriarte, R., and T. Garland. 1996. Testing hypotheses of correlated evolution using phylogenetically independent contrasts: sensitivity to deviations from Brownian motion. Systematic Biology 45:27�47. http://www.biology.ucr.edu/people/faculty/Garland/DiazGa96.pdf Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Bryan McLean [bryansmcl...@gmail.com] Sent: Tuesday, May 10, 2016 1:24 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] simulating continuous data Hi list, I�m working to simulate multiple continuous characters on a known phylogeny (using several of the standard models), and I want to compare properties of the simulated datasets to an empirical dataset. My question is: what is the standard method for ensuring that those datasets (simulated, empirical) are actually directly comparable, i.e. scaled similarly? Does this involve specifying a sensible root state (e.g. ancestral reconstruction) OR just rescaling one or the other datasets before or after the analysis? Forgive me if this is a bit of a naive question, just trying to get a sense of standard practices. -Bryan McLean ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] simulating continuous data
This is a good point and one that is often glossed over. We talked about it quite a bit here: Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265–292. http://www.biology.ucr.edu/people/faculty/Garland/GarlEA93.pdf Surely you want to do various descriptive statistics on your simulated data sets to see how they compare with the real one, and presumably you want some of those to include the phylogenetic versions (e.g., conventional and phylogenetic estimates of the correlation coefficient if you are simulating two traits). I think it is also really important to consider models that have limits to trait evolution (again, see the paper listed above). Those limits can interact strongly with starting (root) values, especially if you include evolutionary trends. Check Figure 1 in this paper: Diaz-Uriarte, R., and T. Garland. 1996. Testing hypotheses of correlated evolution using phylogenetically independent contrasts: sensitivity to deviations from Brownian motion. Systematic Biology 45:27–47. http://www.biology.ucr.edu/people/faculty/Garland/DiazGa96.pdf Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Bryan McLean [bryansmcl...@gmail.com] Sent: Tuesday, May 10, 2016 1:24 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] simulating continuous data Hi list, I’m working to simulate multiple continuous characters on a known phylogeny (using several of the standard models), and I want to compare properties of the simulated datasets to an empirical dataset. My question is: what is the standard method for ensuring that those datasets (simulated, empirical) are actually directly comparable, i.e. scaled similarly? Does this involve specifying a sensible root state (e.g. ancestral reconstruction) OR just rescaling one or the other datasets before or after the analysis? Forgive me if this is a bit of a naive question, just trying to get a sense of standard practices. -Bryan McLean ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Normal distribution in trait values before testing for phylogenetic signals?
These descriptive statistics (e.g., K of Blomberg et al. 2003) and related tests do not require normality of the data under analysis. However, your results may be highly dependent on the type of transform you use (or do not use), so all results must be considered in the context of that transform. In general, I'd be inclined to apply some sort of "sensible" transform, such as log, if I were working with a large number of species that varied widely, such as body masses of mice to elephants. Check the Blomberg et al. (2003) paper for a bit more discussion about this, and to see what they did for the various traits studied. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Ting-Wen Timothy Chen [tch...@gwdg.de] Sent: Wednesday, April 13, 2016 11:02 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Normal distribution in trait values before testing for phylogenetic signals? Hi all, I'm working on the phylogenetic signal tests for continuous traits using K statistics and Pagel's lambda implemented in phytools. I'd like to know whether normal distribution of the trait values must be tested before phylogenetic signals are tested, and if the trait data are not normal distributing, do I have to transform the data? I have looked for this issue in my paper collections as well as on the internet, but still have no idea. It would be very nice if someone can give me some advice, or any paper/reference/website would also be very helpful. Thanks a lot! Best, Ting-Wen -- Ting-Wen Chen J.F. Blumenbach Institute of Zoology and Anthropology Georg August University Goettingen Berliner Str. 28 D-37073 Goettingen, Germany Tel: +49-55139-10943 ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Testing for relationship between one categorical and one continuous variable in a phylogenetic framework.
Alejandro is correct. You can also do it with phylogenetically independent contrasts or computer simulations: Garland Jr., T., P. H. Harvey, and A. R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41:18–32. Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265–292. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Alejandro Gonzalez Voyer [alejandro.gonza...@iecologia.unam.mx] Sent: Friday, April 08, 2016 2:01 PM To: Sean McKenzie Cc: R-phylo Mailing-list Subject: Re: [R-sig-phylo] Testing for relationship between one categorical and one continuous variable in a phylogenetic framework. Hello Sean, If the continuous variable is the “response” and the “independent” variable the discrete one, you can use PGLS, this would be akin to an ANOVA and you can do it accounting for phylogenetic non-independence. Is this what you were after? Cheers Alejandro ___ Dr Alejandro Gonzalez Voyer Laboratorio de Conducta Animal Instituto de Ecología Circuito Exterior S/N Ciudad Universitaria Universidad Nacional Autónoma de México México, D.F. 04510 México Tel: +52 55 5622 9044 E-mail: alejandro.gonza...@iecologia.unam.mx Web: www.alejandrogonzalezvoyer.com > El 08/04/2016, a las 15:56, Sean McKenzie escribió: > > Hello, I have a two traits, one categorical (binary) and one continuous, > and I want to test for a relationship between them accounting for > phylogenetic signal. I have found a plethora of sources for examining > relationships between multiple categorical traits and many others for > examining multiple continuous traits, but I have been hard pressed to find > a test for one categorical and one continuous trait. A random blog post I > stumbled across said I could use either standard phylogenetic independent > contrasts (e.g. pic in ape) or general estimating equations (e.g. > compar.gee in ape). Unfortunately the examples only used continuous data > and tested for significance with regressions through the origin (e.g. > lm(var1_pic ~ var2_pic - 1) or compar.gee(var1 ~ var2 - 1, phy = tree) ). > This seems wrong when one variable was categorical, no? > > So, are PICs and PGEEs really appropriate for a single categorical and a > single continuous variable? If so, what is the appropriate way to test for > significance? If not, or if there's a better way, how can I test this? > > Thanks! > > Sean > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] How to use categorical vectors in package ape for phylogenetic independent contrasts
Thanks for Emmanual.
Kate, I think the original explanation of dummy variables with independent
contrasts is here:
Garland Jr. T., P.H. Harvey, and A.R. Ives. 1992. Procedures for the analysis
of comparative data using phylogenetically independent contrasts. Systematic
Biology 41:18–32.
Cheers,
Ted
Theodore Garland, Jr., Professor
Department of Biology
University of California, Riverside
Riverside, CA 92521
Office Phone: (951) 827-3524
Facsimile: (951) 827-4286 (not confidential)
Email: tgarl...@ucr.edu
http://www.biology.ucr.edu/people/faculty/Garland.html
http://scholar.google.com/citations?hl=en&user=iSSbrhwJ
Director, UCR Institute for the Development of Educational Applications
Editor in Chief, Physiological and Biochemical Zoology
Fail Lab: Episode One
http://testtube.com/faillab/zoochosis-episode-one-evolution
http://www.youtube.com/watch?v=c0msBWyTzU0
From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Emmanuel
Paradis [emmanuel.para...@ird.fr]
Sent: Friday, March 04, 2016 6:30 AM
To: Kate Boyce-Miles; r-sig-phylo@r-project.org
Subject: Re: [R-sig-phylo] How to use categorical vectors in package ape for
phylogenetic independent contrasts
Hi Kate,
You can compute PICs for a categorical variable in the same way than
you enter it in a linear model, that is by first computing its
"contrasts" (this is different from the "P-I-Contrasts", though both
have some conceptual similarities). The easiest way to do it is to use
the function model.matrix. For instance, we generate a tree (tr) and two
variables simulated on that tree, one continuous (x) and one discrete
(z) with three levels:
tr <- rphylo(50, .1, 0)
x <- rTraitCont(tr)
z <- rTraitDisc(tr, k = 3)
We compute the PICs for x the usual way:
pic.x <- pic(x, tr)
And for z:
Z <- model.matrix(~ z)[, -1]
rownames(Z) <- names(z)
pic.z <- apply(Z, 2, pic, phy = tr)
Note that we drop the column with 1's, so Z is a matrix with 2 columns
(number of levels - 1). Thus, pic.z is a matrix with 49 rows (n - 1) and
2 columns. We can now perform a linear regression with the two sets of PICs:
lm(pic.x ~ pic.z - 1)
This should give you the same coefficients than a PGLS like this:
library(nlme)
gls(x ~ z, correlation = corBrownian(phy = tr))
A very nice explanation of this can be found in Blomberg et al. (2012,
Syst. Biol.).
The two columns in pic.z should not be considered separately, in the
same way than in an ANOVA where there are 2 df associated with a 3-level
factor.
HTH
Best,
Emmanuel
Le 04/03/2016 01:07, Kate Boyce-Miles a écrit :
>
> Hello
>
> I have been using numeric vectors to perform phylogenetic independent
> contrasts of several ecological variables for the cat phylogeny, using the
> ape package in r, and I was wondering how this can be done using categorical
> values?
>
> Example of how I have been using numeric values:
>
> tree <- read.newick("phylogeny.txt")
>
> tree <- as.phylo(tree)
>
> tree <- root(tree,1)
>
> x<-c() # x would vectors, for instance rainfall. One value for each
> operational taxonomic unit on the tree, and the same order as they appear in
> the newick file.
>
> pic.rain <- pic(x, tree, scaled = T, var.contrasts = F, rescaled.tree = F) #
> To generate the phylogenetic independent contrasts
>
> cor.test(pic.rain,pic.rain) # To test for a correlation, obviously for rain
> and rain it would be 1.
>
> How can this be done using data such as habitat, i.e grassland/forest/desert,
> or activity, i.e. nocturnal, diurnal etc?
>
> I would be very grateful for any suggestions.
>
> Kate.
>
> [[alternative HTML version deleted]]
>
> ___
> R-sig-phylo mailing list - R-sig-phylo@r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
> Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
>
>
> Pour nous remonter une erreur de filtrage, veuillez vous rendre ici :
> http://f.security-mail.net/301tdFND1Ht
>
>
___
R-sig-phylo mailing list - R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
___
R-sig-phylo mailing list - R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] multiple regression with binomial distribution
You might also want to read about the background, starting with these two papers, both available at my website: Ives, A. R., and T. Garland, Jr. 2010. Phylogenetic logistic regression for binary dependent variables. Systematic Biology 59:9-26. Ives, A. R., and T. Garland, Jr. 2014. Phylogenetic regression for binary dependent variables. Pages 231-261 in Modern Phylogenetic Comparative Methods and Their Application in Evolutionary Biology: Concepts and Practice, L. Z. Garamszegi, ed. Springer: Heidelberg. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Florian Boucher [floflobouc...@gmail.com] Sent: Tuesday, January 26, 2016 8:34 AM To: Knappová Jana Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] multiple regression with binomial distribution Hi Jana, If your data are proportions, you might want to model their logit, ie log(X/(1-X)), rather than their raw value. You can then use PGLS in the same manner as you did. An ML value of 0 for lambda indicates that you are actually fitting a non-phylogenetic model, but it is safer to allow for lambda to vary in case there would be signal in the residuals of the regression. If your data are 0/1, then you should rather use phylogenetic logistic regression. This is implemented for example in the 'phyloglm' function in the R package 'phylolm'. Cheers, Florian 2016-01-26 16:16 GMT+01:00 Knappová Jana : > Hi everyone, > > I am not very familiar with phylogenetics, but i would like to incorporate > it somehow into my work and i appreciate any suggestions. > > Is it possible to use any kind of phylogenetically informed analysis in > case of multiple regression (one response ~ multiple predictors) in case I > assume binomial distribution of response variable? I try to explain species > occurrence (proportional data) by a couple of species traits. > > I tried pgls function from "caper" R package, but I am not sure about it. > Generally, it seems that there is not strong signal in my data, when I used > lambda="ML" than lambda was set to zero. > > Thanks > > Jana Knappová > Botanický ústav AV ČR, v. v. i. | Institute of Botany of the CAS > Zámek 1, 252 43 Průhonice | Zamek 1, CZ-25243 Pruhonice > Česká republika | Czech Republic > jana.knapp...@ibot.cas.cz > www.ibot.cas.cz > www.pruhonickypark.cz > Telefon: 271015401, 737375227 | Phone: +420-271015401, 737375227 > Fax: 271015105 | Fax: +420-271015105 > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at > http://www.mail-archive.com/r-sig-phylo@r-project.org/ -- Florian Boucher Postdoctoral researcher, Institute of Systematic Botany, Zürich [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] PGLS multiple regression with dummy variables and interaction terms
Dear Patrick, Yes, many, many examples could be cited, and I'll add a couple of my own below. More generally, though, remember that there is generally nothing "magic" about phylogenetically informed statistical methods, such as so-called PGLS. PGLS is just GLS done with use of phylogenetic information. And GLS is just OLS with some assumptions relaxed. (Of course, optimal estimation may get a little more complicated in some cases.) And don't forget that PGLS is the same as Felsenstein's (1985) phylogenetically independent contrasts! (Not including more complicated models that estimate optimal transformations of phylogenetic branch lengths.) Garland, Jr., T., and A. R. Ives. 2000. Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. American Naturalist 155:346–364. Gartner, G. E. A., J. W. Hicks, P. R. Manzani, D. V. Andrade, A. S. Abe, T. Wang, S. M. Secor, and T. Garland Jr. 2010. Phylogeny, ecology, and heart position in snakes. Physiological and Biochemical Zoology 83:43–54. Blomberg, S. P., J. G. Lefevre, J. A. Wells, and M. Waterhouse. 2012. Independent contrasts and PGLS regression estimators are equivalent. Systematic Biology 61:382–391. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Chris Organ [organch...@gmail.com] Sent: Saturday, August 08, 2015 8:48 AM To: Patrick Gemmell; r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] PGLS multiple regression with dummy variables and interaction terms Dear Patrick, Sure - see the following as an example (using BayesTraits, but easily done in R too): Organ, C. L., A. Canoville, R. R. Reisz, and M. Laurin. (2011). Paleogenomic data suggest mammal-like genome size in the ancestral amniote and derived large genome size in amphibians. Journal of Evolutionary Biology. 24: 372–380. Best, Chris _ Chris Organ Department of Microbiology and Immunology Department of Earth Sciences Montana State University, Bozeman, MT 59717 or...@montana.edu www.organlab.net On Fri, Aug 7, 2015 at 11:06 AM, Patrick Gemmell wrote: > Dear R-sig-phylo list, > I have a non-phylogenetic multiple regression y ~ x1 + ... + xn + d1 + d2 + > d3, where x1 ... xn are continuous variables and d1, d2 and d3 are dummy > variables (i.e. for each species in the tree I may have up to 4 data points). > In fact, sometimes my dummy variables interact with my continuous variables. > Can I perform a similar multiple regression using the PGLS method? I am very > new to phylogenetic analysis with R, but it seems that both caper and ape > need to assign tip labels to row names, and of course, such row names would > not be unique and therefore make R unhappy. > Any advice would be appreciated. Thank you for your time. > -- Patrick > > > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] simulating rate shifts
Hi Eric, The thing is that it becomes very cumbersome to write user-friendly software with generalized simulation routines to allow various (unlimited?) ad hoc changes in rates of trait evolution, selection regimes (e.g., OU), rates of speciation, etc. so, the expedient approach is often called for. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Eric [eric...@hotmail.com] Sent: Saturday, August 01, 2015 1:12 AM To: Liam J. Revell Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] simulating rate shifts Hi Liam, Yes, this is what I meant. I’m a little surprised that no process-based simulation - rather than prune-and-graft - method exists. In any case, thank you for the example, it is very helpful. Regards, Eric > On Jul 31, 2015, at 8:07 PM, Liam J. Revell wrote: > > Hi Emmanuel & Eric. > > I think it's possible that Eric is interested in simulating shifts in the > lineage diversification (speciation and/or extinction) rate, rather than the > rate of phenotypic trait evolution. > > It is possible to simulate various scenarios of trait-based diversification > in the package diversitree, I believe; however if you are merely interested > in taking (or simulating some tree) and then imagining (and simulating) a > different diversification process for part of that tree, then this is exactly > the same as simulating first one tree under process 1 - then pruning a > subtree, simulating a replacement subtree under process 2, and attaching the > new subtree from whence the previous subtree was removed. > > Since this is easier said then done, I posted a more detailed worked example > on my blog here: > http://blog.phytools.org/2015/07/simulating-arbitrary-shift-in.html. > > All the best, Liam > > Liam J. Revell, Assistant Professor of Biology > University of Massachusetts Boston > web: http://faculty.umb.edu/liam.revell/ > email: liam.rev...@umb.edu > blog: http://blog.phytools.org > > On 7/31/2015 10:57 AM, Emmanuel Paradis wrote: >> Hi Eric, >> >> See the function rTraitCont in ape: the parameters of the BM or OU model >> can be branch-specific, so it's easy to specify a change in parameter(s) >> at a given node. There's an example there: >> >> http://www.mpcm-evolution.org/practice/online-practical-material-chapter-13/chapter-13-2-traits >> >> >> Cheers, >> >> Emmanuel >> >> Le 30/07/2015 11:31, Eric Lewitus a écrit : >>> Hello, >>> >>> There are several functions available for simulating rate shifts in >>> trees (e.g., SimTree, TESS), but these implement tree-wide shifts, >>> which are somewhat unrealistic, rather than shifts descending from a >>> particular node. Is it possible to implement a more realistic rate >>> shifted tree? Has such a thing already been implemented? >>> >>> Thanks. >>>[[alternative HTML version deleted]] >>> >>> ___ >>> R-sig-phylo mailing list - R-sig-phylo@r-project.org >>> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo >>> Searchable archive at >>> http://www.mail-archive.com/r-sig-phylo@r-project.org/ >>> >>> >>> Pour nous remonter une erreur de filtrage, veuillez vous rendre ici : >>> http://f.security-mail.net/301iAWLdbwV >>> >>> >> >> ___ >> R-sig-phylo mailing list - R-sig-phylo@r-project.org >> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo >> Searchable archive at >> http://www.mail-archive.com/r-sig-phylo@r-project.org/ > ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] R-squared alternative for gls
It's not that r squared isn't meaningful for generalized least squares but rather that it cannot be compared directly with values from OLS models. Cheers, Ted From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Monday, July 20, 2015 4:11 AM To: R phylo mailing list mailing list Subject: [R-sig-phylo] R-squared alternative for gls Hello members, I'm creating relative values os semicircular canal size by fitting a PGLS and extracting residuals (using phyl.resid). I've heard that R-squared isn't meaningful in gls models. What I'm trying to do is to know is which of the two independent variables is best and on a lm () model I would just check the R-squared. Is the alternative, in the case of gls(), looking at the AIC or logLik? Thanks in advance. Best regards, Sérgio. -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal ᐧ [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] testing for variation in rates of evolution among traits
If everything is log-transformed then the variance of phylogenetically independent contrasts or, equivalently, the MSE (if I remember correctly) from a PGLS analysis is directly related to the rate of trait evolution. I'm not sure of the best way to test for statistical differences among traits, but I am sure you could do this with simulations. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Karla Shikev [karlashi...@gmail.com] Sent: Thursday, July 16, 2015 2:13 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] testing for variation in rates of evolution among traits Hi there, I've come across several methods to test for differences in the rate of evolution among branches in a tree, but I can't find methods to test for differences in rates of evolution of different traits on the same species (ex. if wing size evolution is faster than than overall body size evolution). Any suggestions? Thanks! Karla [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] PGLS with non-ultrametric tree
Yep, we've all been down this road before, but I'm too old to remember! Thanks and cheers, Ted From: Paolo Piras [paolo.pi...@uniroma3.it] Sent: Thursday, July 16, 2015 2:38 PM To: Solomon Chak; Theodore Garland Jr Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] PGLS with non-ultrametric tree I really think this post could be useful for the discussion https://stat.ethz.ch/pipermail/r-sig-phylo/2010-September/000773.html best paolo Da: R-sig-phylo per conto di Solomon Chak Inviato: giovedì 16 luglio 2015 23.19 A: Theodore Garland Jr Cc: r-sig-phylo@r-project.org Oggetto: Re: [R-sig-phylo] PGLS with non-ultrametric tree Hi Dr. Garland, I got the impression that an ultrametric tree is needed for comparative analysis from this book <http://www.springer.com/us/book/9783662435496>ch. 2 pg. 38: "most comparative analyses assume that the tree is ultrametric, as the majority of analyses deal with evolution of phenotypic traits of extant species with the underlying assumption is that the time available for phenotypic evolution is the same for all taxa". I understand that the methodology of PGLS can use any tree form, I suppose there're different assumptions if I used an additive tree or an ultrametric tree. Could you guide me to some references that explicitly discuss about this? Is it necessary to use an ultrametric tree If I want to estimate lambda in the residual error of the gls and do a branch length transformation as Revell 2010 <http://onlinelibrary.wiley.com/doi/10./j.2041-210X.2010.00044.x/pdf> suggested? Many thanks! Cheers, Solomon Chak --- *Solomon **Tin Chi **Chak* Ph.D. Candidate Tel: (804) 684-7484 Marine Biodiversity Laboratory <http://www.vims.edu/research/units/labgroups/marine_biodiversity/index.php> Virginia Institute of Marine Science <http://www.vims.edu/> College of William and Mary <http://www.wm.edu/> PO Box 1346 / Rt. 1208 Greate Rd., Gloucester Pt, VA 23062, USA On Thu, Jul 16, 2015 at 2:55 PM, Theodore Garland Jr < theodore.garl...@ucr.edu> wrote: > Dear Solomon, > > I don;t know where you are trying to go with this, but any proper > implementation of PGLS (I am not talking about methods that transform the > branch lengths with things like Grafen's rho, Pagel's lambda or OU models) > should be able to use any tree of any shape. This is the same as for > phylogenetically independent contrasts. > > Cheers, > Ted > > Theodore Garland, Jr., Professor > Department of Biology > University of California, Riverside > Riverside, CA 92521 > Office Phone: (951) 827-3524 > Facsimile: (951) 827-4286 (not confidential) > Email: tgarl...@ucr.edu > http://www.biology.ucr.edu/people/faculty/Garland.html > http://scholar.google.com/citations?hl=en&user=iSSbrhwJ > > Director, UCR Institute for the Development of Educational Applications > > Editor in Chief, Physiological and Biochemical Zoology > > Fail Lab: Episode One > http://testtube.com/faillab/zoochosis-episode-one-evolution > http://www.youtube.com/watch?v=c0msBWyTzU0 > > > From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of > Solomon Chak [tc...@vims.edu] > Sent: Thursday, July 16, 2015 3:31 AM > To: r-sig-phylo@r-project.org > Subject: [R-sig-phylo] PGLS with non-ultrametric tree > > Dear all, > > What are the pros and cons of the two methods to do pgls with a > non-ultrametric tree? Many thanks! > > 1) Convert the additive tree to ultrametric using penalized likelihood > (Sanderson 2002) with ape:: chronopl after cross-validation to find the > best lambda. > > 2) Use weighted least-squares in which gls(y ~ x, > correlation=corPagel(value=0.1, phy=phy, fixed=F), > weights=varFixed(~tip.heights), > data=dat) > > Cheers, > Solomon Chak > > --- > *Solomon **Tin Chi **Chak* > Ph.D. Candidate > Tel: (804) 684-7484 > Marine Biodiversity Laboratory > < > http://www.vims.edu/research/units/labgroups/marine_biodiversity/index.php > > > Virginia Institute of Marine Science <http://www.vims.edu/> > College of William and Mary <http://www.wm.edu/> > PO Box 1346 / Rt. 1208 Greate Rd., Gloucester Pt, VA 23062, USA > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at > http://www.mail-archive.com/r-sig-phylo@r-project.org/ > [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-
Re: [R-sig-phylo] PGLS with non-ultrametric tree
Dear Solomon, I don;t know where you are trying to go with this, but any proper implementation of PGLS (I am not talking about methods that transform the branch lengths with things like Grafen's rho, Pagel's lambda or OU models) should be able to use any tree of any shape. This is the same as for phylogenetically independent contrasts. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Solomon Chak [tc...@vims.edu] Sent: Thursday, July 16, 2015 3:31 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] PGLS with non-ultrametric tree Dear all, What are the pros and cons of the two methods to do pgls with a non-ultrametric tree? Many thanks! 1) Convert the additive tree to ultrametric using penalized likelihood (Sanderson 2002) with ape:: chronopl after cross-validation to find the best lambda. 2) Use weighted least-squares in which gls(y ~ x, correlation=corPagel(value=0.1, phy=phy, fixed=F), weights=varFixed(~tip.heights), data=dat) Cheers, Solomon Chak --- *Solomon **Tin Chi **Chak* Ph.D. Candidate Tel: (804) 684-7484 Marine Biodiversity Laboratory <http://www.vims.edu/research/units/labgroups/marine_biodiversity/index.php> Virginia Institute of Marine Science <http://www.vims.edu/> College of William and Mary <http://www.wm.edu/> PO Box 1346 / Rt. 1208 Greate Rd., Gloucester Pt, VA 23062, USA [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] simulating a labile trait
Hi Glenn, I am not up on the latest simulation options in R, but for some relevant general discussions you might check these old papers (PDFs are on my website): Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265-292. Díaz-Uriarte, R., and T. Garland, Jr. 1996. Testing hypotheses of correlated evolution using phylogenetically independent contrasts: sensitivity to deviations from Brownian motion. Systematic Biology 45:27-47. Abstract [PDF file] Díaz-Uriarte, R., and T. Garland, Jr. 1998. Effects of branch length errors on the performance of phylogenetically independent contrasts. Systematic Biology 47:654-672. Garland, T., Jr., and R. Díaz-Uriarte. 1999. Polytomies and phylogenetically independent contrasts: an examination of the bounded degrees of freedom approach. Systematic Biology 48:547-558. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Glenn Seeholzer [seeholzer.gl...@gmail.com] Sent: Friday, July 10, 2015 9:56 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] simulating a labile trait Hi all, I'm looking for a way to simulate a labile trait given a phylogenetic tree of say 200-300 tips. The ultimate goal is to examine the behavior of BAMM when a trait violates the underlying assumption of Brownian motion trait evolution. I'd like the simulated trait to mimic the evolution of a trait like the climatic-niche, in which there is a bounded trait space and distantly related lineages may converge on the same distribution of trait values (think two clades independently diversifying into the same set of diverse environments). In my empirical data, when I divide a tree into independent subclades, climatic-niche evolution within these subclades is best fit by a BM model relative to OU. Let's assume for the moment that BM is a good model for climatic-niche evolution within subclades, yet not across an entire tree. One thought I had was to divide a simulated tree into subclades, then simulate a BM trait for each subclade. This would result in simulated trait values for all tips in the tree and with the desired pattern of convergence among distantly related lineages. However, I'm unsure if the trait values overall would be that different from a BM trait simulated on the entire tree, or if there a more appropriate approach. Thanks in advance for your help. Cheers, Glenn -- Glenn F. Seeholzer Museum of Natural Science Foster Hall 119 LSU, Baton Rouge, LA 70803 215-872-3017 [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] phyres function R package caper
Thank you, Liam! Cheers, Ted From: Liam J. Revell [liam.rev...@umb.edu] Sent: Wednesday, June 24, 2015 10:24 AM To: Theodore Garland Jr; Sergio Ferreira Cardoso; R phylo mailing list mailing list Subject: Re: [R-sig-phylo] phyres function R package caper Hi all. To the original question, you should be able to get these values first using gls(...,correlation=corBrownian(...)) in nlme & then applying residuals to the fitted model returned by gls. For instance, for data frame X with variables x & y, and ultrametric phylogeny tree, you might compute: library(ape) library(nlme) fit<-gls(y~x,data=X,correlation=corBrownian(1,tree)) residuals(fit) (phytools also has a function for this, phyl.resid, but it does exactly the same thing as the code above, and thus there is really no reason to prefer that function - except perhaps to cross-check your result for errors.) With regards to Ted's comment, indeed these are different quantities. Though the fitted coefficients from a contrasts regression should be the same as above, the residuals will be different (and there will be one fewer of them, besides). These residuals, from the contrasts regression, should be phylogenetically independent; however they are not longer associated with species, but with 'contrasts' or nodes in the tree. To obtain these residuals from a contrasts regression you should be able to do something like: X<-X[tree$tip.label,] ## precautionary pic.x<-pic(X[,"x"],tree) pic.y<-pic(X[,"y"],tree) fit<-lm(pic.y~pic.x-1) residuals(fit) There is no particular reason to prefer one set of quantities over the other - it just depends on what subsequent analyses are intended. In the former case, the residuals are associated with species - but these residuals consequently will be phylogenetically correlated & thus the tree needs to be taken into consideration in any subsequent analysis. The latter residuals are phylogenetically independent, but no longer associated with species. (I hate to cite myself, but this is discussed in my paper Revell 2009; Evolution.) I hope this is of some help. All the best, Liam Liam J. Revell, Assistant Professor of Biology University of Massachusetts Boston web: http://faculty.umb.edu/liam.revell/ email: liam.rev...@umb.edu blog: http://blog.phytools.org On 6/24/2015 12:35 PM, Theodore Garland Jr wrote: > Hi All, > > I am going to suggest that when people want any sort of "phylogenetic > residuals" they do some checking on their own to try to verify what, exactly, > they are getting. Here's one check you can do. Compute phylogenetically > independent contrasts for two traits. Perform a regression (through the > origin, of course) of one trait on the other. Save the residuals. Compare > them with the "phylogenetic residuals" you get from some other program that > does a PGLS regression (not with any transformation of the branch lengths). > Let us know what you find! > > Cheers, > Ted > > Theodore Garland, Jr., Professor > Department of Biology > University of California, Riverside > Riverside, CA 92521 > Office Phone: (951) 827-3524 > Facsimile: (951) 827-4286 (not confidential) > Email: tgarl...@ucr.edu > http://www.biology.ucr.edu/people/faculty/Garland.html > http://scholar.google.com/citations?hl=en&user=iSSbrhwJ > > Director, UCR Institute for the Development of Educational Applications > > Editor in Chief, Physiological and Biochemical Zoology > > Fail Lab: Episode One > http://testtube.com/faillab/zoochosis-episode-one-evolution > http://www.youtube.com/watch?v=c0msBWyTzU0 > > > From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio > Ferreira Cardoso [sff.card...@campus.fct.unl.pt] > Sent: Wednesday, June 24, 2015 9:21 AM > To: R phylo mailing list mailing list > Subject: [R-sig-phylo] phyres function R package caper > > Dear all, > > When I try to get a list os phylogenetic residuals using phyres function > from R package I get this message: Error: could not find function "phyres". > Does anyone know how to solve this problem? > >> phyres(fit.gls1) > Error: could not find function "phyres" > > Best regards, > Sérgio. > > -- > Com os melhores cumprimentos, > Sérgio Ferreira Cardoso. > > > > Best regards, > Sérgio Ferreira Cardoso > > > > > MSc. Paleontology candidate > Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa > Geociências - Universidade de Évora > > Lisboa, Portugal > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.et
Re: [R-sig-phylo] phyres function R package caper
Hi All, I am going to suggest that when people want any sort of "phylogenetic residuals" they do some checking on their own to try to verify what, exactly, they are getting. Here's one check you can do. Compute phylogenetically independent contrasts for two traits. Perform a regression (through the origin, of course) of one trait on the other. Save the residuals. Compare them with the "phylogenetic residuals" you get from some other program that does a PGLS regression (not with any transformation of the branch lengths). Let us know what you find! Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Wednesday, June 24, 2015 9:21 AM To: R phylo mailing list mailing list Subject: [R-sig-phylo] phyres function R package caper Dear all, When I try to get a list os phylogenetic residuals using phyres function from R package I get this message: Error: could not find function "phyres". Does anyone know how to solve this problem? > phyres(fit.gls1) Error: could not find function "phyres" Best regards, Sérgio. -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Non-ultrametric tree PGLS
Agreed! Similarly, Pagel's lambda ad Grafen's rho were designed from a purely statistical perspective, whereas OU and ACDC models are motivated by ties to a possible set of biological processes. Cheers, Ted From: Alejandro Gonzalez Voyer [alejandro.gonza...@iecologia.unam.mx] Sent: Thursday, May 14, 2015 8:47 AM To: Theodore Garland Jr Cc: Sergio Ferreira Cardoso; R phylo mailing list mailing list Subject: Re: [R-sig-phylo] Non-ultrametric tree PGLS Hi Sergio, I would add to Ted’s reply that you are not only considering alternative statistical models but that the evolutionary assumptions from the models you are fitting also differ, and you need to keep this in mind when comparing models. Comparison of AICs or any other estimate of goodness of fit must also involve careful consideration of the assumptions of the models you are comparing. In your particular case, the tree with branch lengths set to equal values (all branch lengths = 1) implies different amount of time to evolve for each of your species (in other words the expected variances - diagonal terms in the variance-covariance matrix - differ between the species), and thus you should consider whether such an assumption makes biological sense in your system. Cheers Alejandro ___ Dr Alejandro Gonzalez Voyer Laboratorio de Conducta Animal Instituto de Ecología Circuito Exterior S/N Ciudad Universitaria Universidad Nacional Autónoma de México México, D.F. 04510 México Tel: +52 55 5622 9044 E-mail: alejandro.gonza...@iecologia.unam.mx<mailto:alejandro.gonza...@iecologia.unam.mx> El 14/05/2015, a las 10:39, Theodore Garland Jr mailto:theodore.garl...@ucr.edu>> escribió: Hi Sergio, I am not quite understanding the situation nor why you see a "problem." if I understand correctly, you are considering these five (5) alternative models for some sort of simple or multiple regression: OLS = star phylogeny PGLS with real-time branch lengths (ultrametric) Pagel's lambda with real-time branch lengths (ultrametric) PGLS with all branch legnths equal to 1.0 Pagel's lambda with all branch lengths equal to 1.0 To help decide which model best fits your data, you can look at AIC or for some comparisons do a likelihood ratio test. My experience is that any of the transform models (Pagel's lambda, Grafen's rho, OU in various implementations, ACDC) can sometimes yield really bizarre results when you start with a non-ultrametric tree. You need to be careful and check the REML likelihood surface for multiple peaks, etc. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu<mailto:tgarl...@ucr.edu> http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Thursday, May 14, 2015 8:32 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Non-ultrametric tree PGLS Hello all, I have an ultrametric phylogenetic tree with divergence times as branch lengths. To see if there was a big difference between using these branch lengths and equal (=1) branch lengths I set all lengths to 1 and ran a PGLS. I ran with Lambda transformations and the estimation is that Lambda is superior than 1 (both with ML and REML estimation). I suppose this is a consequence of the tree being non ultrametric. Is there a solution for this problem or should I, in this case, just ran a GLS (Brownian Motion) to avoid the over estimation of the phylogenetic signal? Best regards, Sérgio. -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal ᐧ [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ [[alternative HTML version deleted]] __
Re: [R-sig-phylo] Non-ultrametric tree PGLS
Hi Sergio, I am not quite understanding the situation nor why you see a "problem." if I understand correctly, you are considering these five (5) alternative models for some sort of simple or multiple regression: OLS = star phylogeny PGLS with real-time branch lengths (ultrametric) Pagel's lambda with real-time branch lengths (ultrametric) PGLS with all branch legnths equal to 1.0 Pagel's lambda with all branch lengths equal to 1.0 To help decide which model best fits your data, you can look at AIC or for some comparisons do a likelihood ratio test. My experience is that any of the transform models (Pagel's lambda, Grafen's rho, OU in various implementations, ACDC) can sometimes yield really bizarre results when you start with a non-ultrametric tree. You need to be careful and check the REML likelihood surface for multiple peaks, etc. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Thursday, May 14, 2015 8:32 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Non-ultrametric tree PGLS Hello all, I have an ultrametric phylogenetic tree with divergence times as branch lengths. To see if there was a big difference between using these branch lengths and equal (=1) branch lengths I set all lengths to 1 and ran a PGLS. I ran with Lambda transformations and the estimation is that Lambda is superior than 1 (both with ML and REML estimation). I suppose this is a consequence of the tree being non ultrametric. Is there a solution for this problem or should I, in this case, just ran a GLS (Brownian Motion) to avoid the over estimation of the phylogenetic signal? Best regards, Sérgio. -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal ᐧ [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Weird estimated Lambda values (PGLS)
Do you have a plot of the likelihoods versus the parameter value? Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Friday, May 08, 2015 3:33 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Weird estimated Lambda values (PGLS) Dear all, I performed a PGLS with a tree with divergence times as branch lengths. I used 1 dependent and 1 independent variable. When I estimate the Lambda (in Pagel's method) I get the Maximum Likelihood Lambda ~0 and the Restricted Maximum Likelihood Lambda 0.878054. This is really strange. I was thinking I might have a tree with incorrect branchlengths but I checked the plot of absolute value of contrast vs. standard deviation for both variables and it's ok (according to Garland et al., 1992). With OU transformation I don't have the same problem. However, the p-value of phylogenetic (both with OU or with Brownian Motion PGLS) aren't very different from the ordinary least-squares. Is there any other ordinary procedure/analysis I should make other than the plots I checked to check my tree? (I understand that there is a chance that the characters I am using have influence and maybe the tree doesn't make any difference). Best regards, Sérgio. -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal ᐧ [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] PGLS vs OUwie?
Dear Will, I suggest that you read the Appendix in this paper Lavin et al. (2008) (available on my website) and then the original papers by Butler and King on OUCH, etc. http://www.jstor.org/stable/10.1086/590395 Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of William Gearty [wgea...@stanford.edu] Sent: Tuesday, May 05, 2015 8:39 AM To: r-sig-phylo Subject: [R-sig-phylo] PGLS vs OUwie? Hi all, This may be a very basic question, but I have been racking my brain since my advisor asked me yesterday in our meeting. What are the fundamental differences (if any) between GLS (PGLS) and something like ouch or OUwie? If there is already literature related to this, please feel free to just point me in that direction. Thanks in advance, Will -- William Gearty PhD Student, Paleobiology Department of Geological Sciences Stanford School of Earth, Energy & Environmental Sciences people.stanford.edu/wgearty [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Non normal PGLS results
If you get such a large difference between ML and REML estimation in this sort of situation then probably either (1) something is wrong with the code (bad search algorithm?) or (2) you have something pathological in your tip data set and/or the tree (e.g., some really long singleton branches or really short sister branches). If you see a big outlier in the residuals, then this is a problem - I would not trust the results. How do the residuals look from a regular OLS analysis? Cheers, Ted From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Saturday, April 18, 2015 10:59 AM To: Liam J. Revell Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Non normal PGLS results Hello, Thank you very much for the answer. I assumed PGLS residuals had to be normal (I saw this thread: https://stat.ethz.ch/pipermail/r-sig-phylo/2012-May/002064.html). In fact, what I'm analysing are standardized residuals (standardized by setting the determinant of the covariance matrix equal to one). I thought something was wrong with the residuals because when estimating the Lamba using ML I get a completely different result from estimating it with REML (e.g., REML est: 0.945; M est: ~0). Do you have any idea of wat could be causing this? Once again, thank you for answering. Best regards, Sérgio. ᐧ 2015-04-18 18:22 GMT+01:00 Liam J. Revell : > Hi Sergio. > > We don't expect the residuals from PGLS to be independent draws from a > normal distribution, but multivariate normal with a correlation structure > given by the tree. > > Here I give some more explanation of this on my blog: > http://blog.phytools.org/2013/02/a-comment-on-distribution-of-residuals.html > . > > Let us know if this is helpful. > > All the best, Liam > > Liam J. Revell, Assistant Professor of Biology > University of Massachusetts Boston > web: http://faculty.umb.edu/liam.revell/ > email: liam.rev...@umb.edu > blog: http://blog.phytools.org > > > On 4/18/2015 1:15 PM, Sergio Ferreira Cardoso wrote: > >> Dear all, >> >> I'm performing PGLS's with an ultrametric phylogenetic tree (divergence >> time as branchlengths). I tied Pagel's Lambda transformation, OU >> transformation, regular GLS and OLS, to compare results. There is one >> problem: the residuals of my analyses are not normal. I tried to remove >> big >> outliers but it made things even worse, because without them there are >> even >> more outliers. >> >> So, what should I do? The results are consistent in both 4 tests. But can >> I >> trust the results of the PGLS? Is it particularly bad if PGLS residuals >> aren't normal? Does it critically afect my results? >> >> Thanks in advance. >> >> Best regards, >> Sérgio. >> >> -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] PGLS transformations
In Mesquite, make sure you have the "Branches proportional to lengths" option checked so that the tree you are looking at shows the real branch lengths it has! this is NOT the default option in Mesquite! Cheers, Ted From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Monday, April 13, 2015 10:13 AM To: Emmanuel Paradis Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] PGLS transformations Hello, Thank you both for the help. Emmanuel, so is there a way to see contrasts in R? Reading this paper - Garland, T., Harvey, P. H., & Ives, A. R. (1992). Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology, 41(1), 18-32. - I was aware of the importance of standardizing contrasts and of comparing Absolute value of standard contrat vs Standard deviation of contrast. I've learned to do this in Mesquite but I don't know if R alows this to be done. When I run the GLS I ask R to estimate the rho, so, a priori I never know the rho value. Maybe I'm really really confused, but here is the reason why I tthink something isn't right with my analysis: I built a phylogenetic tree in Mesquite, and based on several works I used million years as branch lengths. It makes sense for me because I'll be using a fossil on my analysis. But I tested the tree before adding the extinct taxa, so to make sure everything was OK when the tree was ultrametric. I noticed that, for example, whatever the independent variable (X) was, the alpha from OU was extremely high (0.999182, for instance). It would always be 0.999. I thought maybe I was using the wrong transformation... That's why I ended up trying to know if I needed to do something prior to transforming and analysing the tree. Thank you very much. Best regards, Sérgio. ᐧ 2015-04-13 17:17 GMT+01:00 Emmanuel Paradis : > Hi Sérgio, > > There is indeed generally a relationship between branch length > transformations and correlation structures. You may check that with the > function vcv2phylo, e.g.: > > > tr <- rcoal(20) > > co <- corGrafen(1, phy = tr) > > ts <- vcv2phylo(vcv(co)) > > all.equal(tr, ts) > [1] FALSE > > all.equal(compute.brlen(tr), ts) > [1] TRUE > > compute.brlen() transforms the branch lengths according to Grafen's model > with parameter rho = 1 (by default). Some other transformations of branch > lengths are available in package geiger. > > Best, > > Emmanuel > > Le 12/04/2015 21:47, Sergio Ferreira Cardoso a écrit : > >> Hi everyone, >> >> I'm relatively new in phylogenetic comparative methods. I'm a little >> confused about branch length transformations. I'm using a tree with >> divergence time (My) as branch lengths. When I use corPagel, corGrafen or >> corMartins in R, the branch lengths, are the branch lengths automatically >> transformed? e.g., gr.mammals<-corGrafen(1,phylo,fixed=F); >> fit<-gls(FCL~logBodymass,correlation=gr.mammals,data=df,method="ML"). >> My question may sound a bit nonsense but I've seen in some papers (e.g., >> Spoor, >> F., Garland, T., Krovitz, G., Ryan, T. M., Silcox, M. T., & Walker, A. >> (2007). The primate semicircular canal system and locomotion. *Proceedings >> of the National Academy of Sciences*, *104*(26), 10808-10812.) the >> indication that a PGLS was made without branch transformation, but no >> reference is made to the model (maybe it's corBrownian). >> >> Thank you very much. >> >> Best regards, >> Sérgio. >> >> > -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] PGLS transformations
Dear Sergio, I do not use R very much, relying instead on our Matlab programs that accompany Lavin et al. (2008): Lavin S.R., W.H. Karasov, A.R. Ives, K.M. Middleton, and T. Garland Jr. 2008. Morphometrics of the avian small intestine compared with that of nonflying mammals: a phylogenetic approach. Physiological and Biochemical Zoology 81:526–550. I would strongly suggest that you read the appendix to this paper regarding methods, their history, and terminology. Note that a GLS model does not inherently use any kind of branch length transformation. It uses the tree as inputted. It is mathematically equivalent to Felsenstein's (1985) phylogenetically independent contrasts. It implicitly assumes character evolution (or residual character evolution) akin to Brownian motion. Once you start estimating some branch lengths transformation along with regression parameters, you have a different beast. See the appendix noted above and, for the original source of this sort of model, Grafen (1989). Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Sergio Ferreira Cardoso [sff.card...@campus.fct.unl.pt] Sent: Sunday, April 12, 2015 12:47 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] PGLS transformations Hi everyone, I'm relatively new in phylogenetic comparative methods. I'm a little confused about branch length transformations. I'm using a tree with divergence time (My) as branch lengths. When I use corPagel, corGrafen or corMartins in R, the branch lengths, are the branch lengths automatically transformed? e.g., gr.mammals<-corGrafen(1,phylo,fixed=F); fit<-gls(FCL~logBodymass,correlation=gr.mammals,data=df,method="ML"). My question may sound a bit nonsense but I've seen in some papers (e.g., Spoor, F., Garland, T., Krovitz, G., Ryan, T. M., Silcox, M. T., & Walker, A. (2007). The primate semicircular canal system and locomotion. *Proceedings of the National Academy of Sciences*, *104*(26), 10808-10812.) the indication that a PGLS was made without branch transformation, but no reference is made to the model (maybe it's corBrownian). Thank you very much. Best regards, Sérgio. -- Com os melhores cumprimentos, Sérgio Ferreira Cardoso. Best regards, Sérgio Ferreira Cardoso MSc. Paleontology candidate Departamento de Ciências da Terra - FCT /Universidade Nova de Lisboa Geociências - Universidade de Évora Lisboa, Portugal [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] phytools - evaluating significance of pgls.Ives
Andrea, remember that you can and should also do the "OLS" models (i.e., assuming a star phylogeny) with measurement error considered. That's in the programs that accompany Ives, Midford, and Garland (2007, Syst. Biol. 56:252–270), and were in the batch I just sent you. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: R-sig-phylo [r-sig-phylo-boun...@r-project.org] on behalf of Andrea Berardi [andrea.bera...@colorado.edu] Sent: Monday, March 02, 2015 3:57 PM To: r-sig-phylo@r-project.org Cc: Anthony R Ives; Peter Smits Subject: Re: [R-sig-phylo] phytools - evaluating significance of pgls.Ives Thank you all very much for the comments! They are so helpful. Yes, I do only have 8 species, and 3 replicates each. It is not ideal, but it's what we have and we have a phylogeny, so I'd like to try some tests incorporating phylogeny. I probably should have added that I plan on running OLS to test each regression without the tree. This should give me an idea of the general relationship. Regarding Liam's comment in pgls.Ives: Is the "lower=c(1e-8,1e-8)" call of the pgls.Ives call the part where it constrains the slope to (almost)zero? I'll definitely give the MERegPHYSIGv2.m method a try as well, with Tony's diagnostic suggestions. I will also go the LRT route with the data. Thanks very much again for your help! Andrea ~~ Andrea Berardi, PhD Postdoctoral Researcher, Smith Lab EBIO, University of Colorado-Boulder andrea.bera...@colorado.edu On Mar 1, 2015, at 8:42 PM, Liam J. Revell wrote: > Hi Andrea. > > This is not presently implemented, but since this is a likelihood method it > would be straightforward to constrain to a slope of zero and then do a LR > test. This would be probably be the easiest way to test a hypothesis about > the regression. > > That being said, as noted in the function documentation, some problems have > been reported with the optimization algorithm for this model, which is simple > and thus may fail to find the ML solution. Consequently, I would encourage > you to look for other implementations of the method so that you can be > confident in your result. I'm not aware of one in R at this time. > > All the best, Liam > > Liam J. Revell, Assistant Professor of Biology > University of Massachusetts Boston > web: http://faculty.umb.edu/liam.revell/ > email: liam.rev...@umb.edu > blog: http://blog.phytools.org > > On 3/1/2015 10:31 PM, Andrea Berardi wrote: >> Hi all, >> >> I'm just learning how to do PGLS analyses, and I'm looking for advice on how >> to evaluate the significance of the regression fit using pgls.Ives in the >> phytools package. I'm using this function because it incorporates sampling >> error of species means, and my data has about 3 individuals per species, >> with 8 species. My goal is to test whether a flower trait predicts the leaf >> trait, while controlling for shared ancestry. Here is the output from >> pgls.Ives: >> >>> fit <- pgls.Ives(Tree, Flower_trait, Leaf_trait) >>> fit >> $beta >> [1] 96.3963098 0.1292656 >> >> $sig2x >> [1] 22218901073 >> >> $sig2y >> [1] 23027587 >> >> $a >> [1] -10063.150 -1204.422 >> >> $logL >> [1] -158.2337 >> >> $convergence >> [1] 0 >> >> $message >> [1] "CONVERGENCE: REL_REDUCTION_OF_F <= FACTR*EPSMCH" >> >> I am also running pgls on species averages for the traits using the gls >> function in nlme and the corBrownian and corMartins functions in ape. But, >> we are interested in incorporating the within-species variation in our small >> dataset. >> >> Any suggestions would be welcome! >> >> Thanks for your help, >> Andrea >> >> ~~ >> Andrea Berardi, PhD >> Postdoctoral Researcher, Smith Lab >> EBIO, University of Colorado-Boulder >> >> ___ >> R-sig-phylo mailing list - R-sig-phylo@r-project.org >> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo >> Searchable archive at http://www.mail-archive.com/
Re: [R-sig-phylo] How to test if the slope is different from 1 in PGLS?
Just remember that if you have measurement error in the X variable then you likely underestimate the true slope: Ives, A. R., P. E. Midford, and T. Garland. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology 56:252–270. Also, if you have soft polytomies then you may want to subtract some d.f.: Purvis, A., and T. Garland, Jr. 1993. Polytomies in comparative analyses of continuous characters. Systematic Biology 42:569–575. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Xavier Prudent [prudentxav...@gmail.com] Sent: Thursday, August 21, 2014 3:06 AM To: Gustavo Paterno Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] How to test if the slope is different from 1 in PGLS? Hi Gustavo, GLS returns the estimated slope and its uncertainty, you can then compute a t-value where your null-hypothesis is "slope=1", and from that t-value get a p-value. Cheers, Xavier 2014-08-20 21:30 GMT+02:00 Gustavo Paterno : > Dear all, > I am working with flower allometry and want to use PGLS to analyse how the > male and female biomass of the flower scale with total flower biomass. > So simple linear regressions, but log-trasformed. > > my model is: > mod.male <- pgls(male~total, data=flower.data,lambda="ML") > > Besides calculating the slope of the regression I am also interested to > test if the slope is different then 1. How can I do this in PGLS? > I would be very glad if any one could help me with that. > > Kind regards, > Gustavo Paterno > - > Gustavo B. Paterno > Doutorando em Ecologia > Departamento de Ecologia > Universidade Federal do Rio Grande do Norte > Natal, Brasil > - > skype: gustavopaterno > pater...@cb.ufrn.br > > > > > > > > > > > > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at > http://www.mail-archive.com/r-sig-phylo@r-project.org/ > -- *---Xavier Prudent* *Computational biology and evolutionary genomics* *Guest scientist at the Max-Planck-Institut für Physik komplexer Systeme* *(MPI-PKS)* *Noethnitzer Str. 38* *01187 Dresden * *Max Planck-Institute for Molecular Cell Biology and Genetics* *(MPI-CBG)* *Pfotenhauerstraße 108 * *01307 Dresden* *Phone: +49 351 210-2621* *Mail: prudent [ at ] mpi-cbg.de <http://mpi-cbg.de>* *---* [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] multiple traits measured within species and populations
Dear Jorge, For simplicity, I'd start with these two papers, both available on my webpage: Purvis, A., and T. Garland, Jr. 1993. Polytomies in comparative analyses of continuous characters. Systematic Biology 42:569–575. Theodore Garland, Jr., Paul H. Harvey, and Anthony R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41:18–32. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications Editor in Chief, Physiological and Biochemical Zoology Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Jorge Sánchez Gutiérrez [jorgesgutier...@unex.es] Sent: Wednesday, August 20, 2014 9:49 AM To: r-sig-phylo@r-project.org Mailing-list Subject: [R-sig-phylo] multiple traits measured within species and populations Dear All, I'm working on some comparative analyses involving up to 312 populations of birds representing 112 species, one dependent variable (sometimes count data, e.g. species richness), various continuous independent variables (e.g. latitude, body mass), and a categorical variable (habitat: 3 levels). I'd be very grateful if I could get some input into these analyses: I constructed a phylogeny for the species in my data sets. Because some species had been examined in multiple study locations (up to 14), I represented these with multiple points and added them as polytomies to the appropriate species tip in the original tree (as in , e.g., Anderson et al.; Ecology Letters, (2005) 8: 310–318). This yielded an ultrametric tree containing multiple polytomies. I've been running regressions in APE using gls and the correlation structures implemented in the package and it seems to work well except for 'corPagel' when fitted with maximum likelihood (ML). In such a case I get the message: "Error in corFactor.corStruct(object) : NA/NaN/Inf in foreign function call (arg 1)". Example: m1 <- gls(richness ~ Mass + Habitat + Latitude, data= richness, correlation = CorPagel, method = "ML") But most importantly, I'm not sure if treating these points (means of 'populations') as statistically independent replicates is correct. I've read some papers regarding how to account for intraspecific variation in phylogenetic analyses (including those referred to in the 2nd edition of Paradis) and some related posts. However, as far as I know I think these methods account for between-individual variation instead of for between-population variation. In my data set the repeated observations per species do not correspond to different individuals but populations within species. Maybe I'm on the wrong track with this, so any help is very welcome. Best, Jorge Jorge S. Gutiérrez Department of Anatomy, Cell Biology & Zoology. Faculty of Sciences. University of Extremadura. Avenida de Elvas s/n. E-06006 Badajoz, Spain ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Concentrated changes test
Isn't it in Mesquite? Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications http://idea.ucr.edu/ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Matt Pennell [mwpenn...@gmail.com] Sent: Friday, May 16, 2014 9:04 AM To: R-phylo Mailing-list Subject: [R-sig-phylo] Concentrated changes test Hi everyone, Just a quick question: I was wondering if there is a R implementation of Wayne Maddison's (1990) concentrated changes test ( http://www.jstor.org/stable/2409434). After a little looking around, I can't seem to find it but was hoping someone else might have some ideas. Thanks in advance, Matt [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Cross-validation with independent contrasts
Couldn't you also just do this back at the level of the original tree and tip data, creating subsets by pruning the tree before you compute contrasts? Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Director, UCR Institute for the Development of Educational Applications http://idea.ucr.edu/ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Diego Bilski [diegobil...@gmail.com] Sent: Friday, May 16, 2014 7:55 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Cross-validation with independent contrasts Dear all, I'm wondering if it would be statistically/philosophically correct to use n-fold cross-validation to evaluate a linear regression with independent contrasts. My doubt comes from the fact that when simply dividing the IC dataset in, lets say, 10 folds, some folds will remove the contrasts of internal nodes without necessarily removing an entire clade above that point, producing what can be viewed as two independent clades (a graphical example would be, in Felsenstein's seminal paper, fig. 8, remove the contrast at node 13, while keeping those at nodes 9 and/or 10). Any thoughts? Best wishes, Diego Bilski [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] fitting cuadratic model using gnls
I don't know, but one thing to watch out for is a high correlation between X and Xsquared. That gives multicollinearity and can lead to numerical problems. One thing that is commonly done is: 1. standardize the X variable (subtract mean and divide by standard deviation) 2. square that value and use it for the quadratic This is referred to as an "orthogonol polynomial." Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Jorge Sánchez Gutiérrez [jorgesgutier...@unex.es] Sent: Tuesday, December 10, 2013 9:48 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] fitting cuadratic model using gnls Hello everyone, I’m trying to fit nonlinear models using the function “gnls” in ape. For example, I run a simple quadratic regression of moult duration (duration) and latitude (latitude_used) as follows: mod <- duration ~ latitude_used +I(latitude_used^2) init<-list(alpha=1,beta=1) m3 <- gnls(mod, data=data, start=init, correlation = dataCor) However, I get the following error message: “Error in gnls(mod, data=data, start = init, correlation = dataCor) : step halving factor reduced below minimum in NLS step” I’ve also tried with several correlation structures and different starting values for estimation, but without success. I would be very grateful if you can provide any suggestions. Thanks in advance, Jorge S. Gutiérrez [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] best fit vs normality of residuals
Good advice! Cheers, Ted From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Luke Matthews [lmatth...@activatenetworks.net] Sent: Tuesday, December 03, 2013 8:29 AM To: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] best fit vs normality of residuals Hi Agus, If I understand your post correctly, you implemented the two models with exactly the same formula, and phylogenetic tree, and varied only the transform applied to the variables. In one case the transform was to 'scale and center' both the dependent and independent variables while in the other case you rank ordered the all variables. Please clarify if I have this wrong. In this case you would have to compare the normality of the residuals, as when you transform the dependent variable in different ways I think the likelihoods will necessarily shift just because the distribution of the data being explained is shifted. I don't think you can compare likelihoods, AICs, or BICs in this way when you alter the dependent, just as you can't compare likelihood-based values for models that include mostly the same data points but also some different ones. You can only compare likelihood based values for models that have the exact same dependent data but differ in the formulation of the model's independent variables, autocorrelation parameters, etc. As another point, although ranking seemed to fix your normality problem in this case, it may be introducing some other issues. Ranked variables are usually too flat for linear regression analysis, and perhaps the test of residual normality you are using only detects some deviations like skewness (I'm not familiar with the Liliefors test). I would recommend that instead of ranking you plot the distribution of your dependent and independent variables and try the transforms (log, arcsine, etc.) that a standard stats book would recommend given the appearance of each distribution. Some variables may not need transforming at all. It may be just one or two that is causing the problem in the model residuals, in which case you should transform only those variables. Remember that since it's the normality of the residuals that matters, it may be more appropriately fixed by transforming an independent variable than a dependent one. If the dependent itself is normal, then nonnormal! ity might be introduced in the residuals because of the distribution of a particular independent variable. Best Luke Luke J. Matthews | Senior Scientific Director | Activate Networks -- Message: 2 Date: Fri, 29 Nov 2013 12:29:06 -0200 From: Agus Camacho To: "r-sig-phylo@r-project.org" Subject: [R-sig-phylo] best fit vs normality of residuals Message-ID: Content-Type: text/plain Dear colleagues, Im having difficulties to decide whether I choose a phylogenetic GLS model with a higher fit (lower AIC and BIC), or a model in which normality of the residuals, after accounting for phylogenetic signal, is compromised. The number of species is reasonably high (87), but i dont know if that would justify for allowing a highly significant deviation of normality. When using scaled and centered data, i get: AIC BIC logLik 255.2029505 269.5696455 -121.6014753 Correlation Structure: corPagel Formula: ~1 Parameter estimate(s): lambda -0.03313647856 Coefficients: Value Std.Error t-value p-value (Intercept)0.0432999895 0.06632562733 0.6528395024 0.5157 X 0.0425258358 0.03552018760 1.1972300459 0.2347 X2 0.4620358585 0.18471478739 2.5013474287 0.0144 X1:X2 -0.1211398020 0.04969892007 -2.4374735269 0.0170 Liliefors test (thanks Liam for posting on this) gave me: D = 0.1815, p-value = 2.558e-07 I ranked both, the response variable and the factors. My variables had some zeros and in some cases negative values, so thought that would be the simplest and most robust way. But i might be wrong. When ranking all variables: AIC BIC logLik 766.7826784 781.1493734 -377.3913392 Correlation Structure: corPagel Formula: ~1 Parameter estimate(s): lambda 0.1434096557 Coefficients: Value Std.Error t-value p-value (Intercept) 5.615576688 9.195445483 0.610691097 0.5431 X1 0.477054032 0.200882571 2.374790556 0.0199 X2 0.771914482 0.208616720 3.700156356 0.0004 X1:x2 -0.007371999 0.004035148 -1.826946400 0.0714 Lilliefors (Kolmogorov-Smirnov) normality test data: chol(solve(vcv(tree))) %*% residuals(M2) D = 0.0545, p-value = 0.7709 Would anybody have a hint on this? Gracias! Agus -- Agust?n Camacho Guerrero. Doutor em Zoologia. Laborat?rio de Herpetologia, Departamento de Zoologia, Insti
Re: [R-sig-phylo] Transforming data for OU model
I had been imagining that precipitation was a dependent variable, not that it matters. In any case, you have the need for a lower limit of zero and probably an upper limit that is somewhat above the greatest recorded values. I can't see why you would want an OU model for precipitation, though. As Liam wrote, we all need more info about what you are doing. And we need to get the hell off the computer on a holiday! Cheers, Ted From: Liam J. Revell [liam.rev...@umb.edu] Sent: Thursday, November 28, 2013 12:16 PM To: Theodore Garland Jr; Anna Rice; r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Transforming data for OU model Hi Ted & Anna. phytools indeed includes a function for simulating Brownian evolution within bounds. This is in the function fastBM. It is also possible to simulate with a trend in fastBM, or under OU (as of the latest non-CRAN phytools, http://www.phytools.org) - but combining a trend or bounds with OU is not permitted. One question I have is what do you mean when you are saying you are "running an OU model" on precipitation data? Are you fitting a linear regression using GLS with a phylogenetic error structure based on the OU model in which precipitation is a predictor? If so, then it may not matter because multivariate normality of the predictors is not, strictly speaking, an assumption of GLS. Hope that may be helpful. All the best, Liam Liam J. Revell, Assistant Professor of Biology University of Massachusetts Boston web: http://faculty.umb.edu/liam.revell/ email: liam.rev...@umb.edu blog: http://blog.phytools.org On 11/28/2013 1:52 PM, Theodore Garland Jr wrote: > Given the nature of precipitation data, I don't think a log transform makes > sense biologically. It is very different from, say, body mass. > I think you need to simulate with limits, as described originally here: > > Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. > Phylogenetic analysis of covariance by computer simulation. Systematic > Biology 42:265-292. > > Limits can be implemented with or without other things, such as OU and/or a > trend. > > Our programs are in DOS (What's DOS? The last operating system that worked > ...). > However, I think you can also do this in R somewhere, maybe phytools? Liam > Revell, want to jump in here? > > Happy thanksgiving, > Ted > > Theodore Garland, Jr., Professor > Department of Biology > University of California, Riverside > Riverside, CA 92521 > Office Phone: (951) 827-3524 > Wet Lab Phone: (951) 827-5724 > Dry Lab Phone: (951) 827-4026 > Home Phone: (951) 328-0820 > Skype: theodoregarland > Facsimile: (951) 827-4286 = Dept. office (not confidential) > Email: tgarl...@ucr.edu > http://www.biology.ucr.edu/people/faculty/Garland.html > http://scholar.google.com/citations?hl=en&user=iSSbrhwJ > > Inquiry-based Middle School Lesson Plan: > "Born to Run: Artificial Selection Lab" > http://www.indiana.edu/~ensiweb/lessons/BornToRun.html > > Fail Lab: Episode One > http://testtube.com/faillab/zoochosis-episode-one-evolution > http://www.youtube.com/watch?v=c0msBWyTzU0 > > > From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] > on behalf of Anna Rice [annarice...@gmail.com] > Sent: Thursday, November 28, 2013 6:47 AM > To: r-sig-phylo@r-project.org > Subject: [R-sig-phylo] Transforming data for OU model > > Hi all, > > I am running an OU model on data that is constrained by zero (precipitation > values). Therefore I thought to log transform the data. However I still > have some issues with the transformation: > > 1. When all values in the data are larger than 1.0, the log transformation > will still result in all values being positive. > 2. In case I have a value of zero, log transforming it will give -Inf. > > I would greatly appreciate any advice regarding the best way of > transforming and dealing with this kind of data. > > Thanks you, > Anna > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ > ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Transforming data for OU model
Good point, Joe. However, if they have graphics screens (as in our PDTREE.EXE and PDSIMUL.EXE) programs, then you sometimes have trouble with the display of those particular screens. Plain text screens (ASCII characters only) aren't a problem. Also, you can get idiosyncratic problem, such as a laptop that shows the graphics screen just fine, but an external monitor glitches out. Anyway, if anybody wants these "classic" DOS programs, just let me know! Hasta, Ted From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Joe Felsenstein [j...@gs.washington.edu] Sent: Thursday, November 28, 2013 1:38 PM To: r-sig-phylo@r-project.org Mailing-list Subject: Re: [R-sig-phylo] Transforming data for OU model Ted Garland wrote: > Our programs are in DOS (What's DOS? The last operating system > that worked ...). > However, I think you can also do this in R somewhere, maybe > phytools? Liam Revell, want to jump in here? Just to remind people: DOS (also called MSDOS) executable programs can be run in Windows using the Command Prompt tool, which you will find in the Accessories menu that is found in the menu opened by the All Programs tab in the Start Menu. Joe Joe Felsenstein, j...@gs.washington.edu Dept. of Genome Sciences, Univ. of Washington Box 355065, Seattle, WA 98195-5065 USA ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Transforming data for OU model
Given the nature of precipitation data, I don't think a log transform makes sense biologically. It is very different from, say, body mass. I think you need to simulate with limits, as described originally here: Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265-292. Limits can be implemented with or without other things, such as OU and/or a trend. Our programs are in DOS (What's DOS? The last operating system that worked ...). However, I think you can also do this in R somewhere, maybe phytools? Liam Revell, want to jump in here? Happy thanksgiving, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html Fail Lab: Episode One http://testtube.com/faillab/zoochosis-episode-one-evolution http://www.youtube.com/watch?v=c0msBWyTzU0 From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Anna Rice [annarice...@gmail.com] Sent: Thursday, November 28, 2013 6:47 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Transforming data for OU model Hi all, I am running an OU model on data that is constrained by zero (precipitation values). Therefore I thought to log transform the data. However I still have some issues with the transformation: 1. When all values in the data are larger than 1.0, the log transformation will still result in all values being positive. 2. In case I have a value of zero, log transforming it will give -Inf. I would greatly appreciate any advice regarding the best way of transforming and dealing with this kind of data. Thanks you, Anna [[alternative HTML version deleted]] ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] zero length terminal branches in pgls
Another suggestion would be to abandon your "real" molecular branch lengths and use arbitrary ones. In such cases, I generally try three sets of arbitrary branch lengths: 1. all = 1 2. Pagel's (1992) arbitrary 3. Grafen's (1989) arbitrary See here: Garland, T., Jr., P. H. Harvey, and A. R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41:18-32. For #2 and #3, I would also try transformations, as in an OU model. See here, especially the appendix (RegOU, but also RegPagel, RegGrafen): Lavin, S. R., W. H. Karasov, A. R. Ives, K. M. Middleton, and T. Garland, Jr. 2008. Morphometrics of the avian small intestine, compared with non-flying mammals: A phylogenetic perspective. Physiological and Biochemical Zoology 81:526-550. [provides Matlab Regressionv2.m, released as part of the PHYSIG package] We have been using lilkelihoods, AIC or AICc to figure out which model(s) best fit your data (including choosing among sets of independent variables), e.g.: Oufiero, C. E., S. C. Adolph, G. E. A. Gartner, and T. Garland, Jr. 2011. Latitudinal and climatic variation in body size and dorsal scale rows in Sceloporus lizards: a phylogenetic perspective. Evolution 65:3590-3607. Nobody knows how well a "model selection" approach that really works in this specific context, but we are going with it for now until somebody does massive simulations! Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Tom Kraft [thomas.s.kraft...@dartmouth.edu] Sent: Wednesday, July 31, 2013 9:57 AM To: Liam J. Revell Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] zero length terminal branches in pgls Hi Liam, Thank you for your rapid and helpful comments (I really appreciate your including the IC example to help me understand the nature of the problem). Just to clarify your point about using the method by Ives et al. (2007), my understanding is that this would essentially involve combining my taxa that are separated by 0 branch length distance (since they cannot be separated in the phylogeny as distinct species anyways) and using a combination of trait values. My question then is whether it is even worth attempting this if my trait values are substantially different and the most likely explanation for the 0 length branches is the need for more molecular data. In addition, should I be throw out both members of a zero length terminal branch pair, or am I justified in retaining one of the pair randomly? Best, Tom On Wed, Jul 31, 2013 at 12:31 PM, Liam J. Revell wrote: > Hi Tom. > > Yes, for zero length terminal branches it is inadvisable to add a "tiny > amount" to the terminal edge so that the analysis "works". An easy way > to think about this is in terms of Felsenstein's contrasts method (which > is a special case of PGLS). Contrasts are standardized to have the same > expected variance by dividing by the square-root of the subtending > edges. If you set some (terminal) branches to be very small - then the > corresponding contrast will be very large. (Going to Inf as the edge > lengths go to zero!) This will give this contrast very high weight in > your regression. > > Part of the problem here stems from the different effective meaning of a > zero length terminal edge in molecular phylogenetics vs. comparative > biology. In the former - a zero length terminal edge probably means that > we don't have enough sequence data, and thus failed to sample any > substitutions along that edge. Depending on the amount of data that you > have, the edge could be 1000s to millions of years long. A zero length > terminal edge in comparative inference means that the tree ends at a > speciation event - no time elapsed between the speciation event that > created a lineage and the present time. Under those circumstances, we > should expect to find no phenotypic difference between species separated > by zero time! (And it makes sense that comparative methods wouldn't like > data that suggested otherwise.) > > The best solution is to get more data to infer your tree. Failing that, > you could assume that there is the eq
Re: [R-sig-phylo] PGLS vs lm
Hi Tom, So far I have resisted jumping in here, but maybe this will help. Come up with a model for how you think your traits of interest might evolve together in a correlated fashion along a phylogenetic tree. Now implement it in a computer simulation along a phylogenetic tree. Also implement the model with no correlation between the traits. Analyze the data with whatever methods you choose. Check the Type I error rate and then the power of each method. Also check the bias and means squared error for the parameter you are trying to estimate. See what method works best. Use that method for your data if you have some confidence that the model you used to simulate trait evolution is reasonable, based on your understanding (and intuition) about the biology involved. Lots of us have done this sort of thing, e.g., check this: Martins, E. P., and T. Garland, Jr. 1991. Phylogenetic analyses of the correlated evolution of continuous characters: a simulation study. Evolution 45:534-557. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Tom Schoenemann [t...@indiana.edu] Sent: Friday, July 26, 2013 3:21 PM To: Tom Schoenemann Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] PGLS vs lm OK, so I haven't gotten any responses that convince me that PGLS isn't biologically suspect. At the risk of thinking out loud to myself here, I wonder if my finding might have to do with the method detecting phylogenetic signal in the error (residuals?): From: Revell, L. J. (2010). Phylogenetic signal and linear regression on species data. Methods in Ecology and Evolution, 1(4), 319-329. I note the following: "...the suitability of a phylogenetic regression should actually be diagnosed by estimating phylogenetic signal in the residual deviations of Y given our predictors (X1, X2, etc.)." Let's say one variable, "A", has a strong evolutionary signal, but the other, variable "B", does not. Would we expect this to affect a PGLS differently if we use A to predict B, vs. using B to predict A? If so, it would explain my findings. However, given the difference, I can have no confidence that there is, or is not, a significant covariance between A and B independent of phylogeny. Doesn't this finding call into question the method itself? More directly, how is one to interpret such a finding? Is there, or is there not, a significant biological association? -Tom On Jul 21, 2013, at 11:47 PM, Tom Schoenemann wrote: > Thanks Liam, > > A couple of questions: > > How does one do a hypothesis test on a regression, controlling for phylogeny, > if not using PGLS as I am doing? I realize one could use independent > contrasts, though I was led to believe that is equivalent to a PGLS with > lambda = 1. > > I take it from what you wrote that the PGLS in caper does a ML of lambda only > on y, when doing the regression? Isn't this patently wrong, biologically > speaking? Phylogenetic effects could have been operating on both x and y - we > can't assume that it would only be relevant to y. Shouldn't phylogenetic > methods account for both? > > You say you aren't sure it is a good idea to jointly optimize lambda for x & > y. Can you expand on this? What would be a better solution (if there is > one)? > > Am I wrong that it makes no evolutionary biological sense to use a method > that gives different estimates of the probability of a relationship based on > the direction in which one looks at the relationship? Doesn't the fact that > the method gives different answers in this way invalidate the method for > taking phylogeny into account when assessing relationships among biological > taxa? How could it be biologically meaningful for phylogeny to have a > greater influence when x is predicting y, than when y is predicting x? Maybe > I'm missing something here. > > -Tom > > > On Jul 21, 2013, at 8:59 PM, Liam J. Revell wrote: > >> Hi Tom. >> >> Joe pointed out that if we assume that our variables are multivariate >> normal, then a hypothesis test on the regression is the same as a test that >> cov(x,y) is
Re: [R-sig-phylo] Problem with fitContinuous function
I just want to emphasize thism point: "one other thing that is useful, but which we don't do enough, is to simply visualize the results" Good old fashioned phylogenetically independent contrasts (Felsenstein, 1985; Garland et al., 1992) are really useful for this, at least with your original tree. You can often spot influential contrasts, then try to track back and see if it is a likely problem with the tip data, the branch lengths (too short for that contrast?) or both. Sometimes, of course, a big contrast is real and reflects unusally rapid evolutionary divergence in the trait in that bifurcation of your tree. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Brian O'Meara [bome...@utk.edu] Sent: Wednesday, July 17, 2013 10:52 AM To: Slater, Graham Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Problem with fitContinuous function In addition to Liam's and Graham's suggestions, one other thing that is useful, but which we don't do enough, is to simply visualize the results. Do plot(ouTree(phy, your_estimate)) and you can see what Graham means about signal. Do phytools' contMap with and without the transformed tree to look at how your traits are distributed: maybe you have one extreme value that's driving this? Are your tree's starting branch lengths very weird? (ultrametric trees aren't required, but often make a lot of sense for this sort of analysis for trees of extant organisms). There may be an issue with tree or data that becomes obvious with this sort of inspection, or you may figure out that a new kind of model is needed (maybe your data suggest a two mean OU model). You don't want to make post hoc hypotheses when doing this, of course, but you can see problems this way. Best, Brian ___ Brian O'Meara Assistant Professor Dept. of Ecology & Evolutionary Biology U. of Tennessee, Knoxville http://www.brianomeara.info Students wanted: Applications due Dec. 15, annually Postdoc collaborators wanted: Check NIMBioS' website Calendar: http://www.brianomeara.info/calendars/omeara On Wed, Jul 17, 2013 at 1:26 PM, Slater, Graham wrote: > Hi Oscar, > > Liam is of course correct that you can change the bounds. The default > bounds in geiger are quite wide however. It's probably worth noting that > the fact that both OU and trend (=linear rate change through time) have ML > parameter estimates at their optimization bounds suggests that there is > very little 'phylogenetic signal' in your trait. This is because both OU > and increasing rates through time result in trait values for closely > related tips in ultrametric trees that are much more dissimilar than > expected in high signal models, like BM or ACDC/trend with declining rates > through time. It's definitely worth increasing the bounds for optimization > but I'd also suggest performing some 'phylogenetic signal' tests to find > out whether you are able to say much about the processes/rates underlying > the evolution of your trait > > Cheers, > > Graham > --- > -- > Graham Slater > Peter Buck Post-Doctoral Fellow > Department of Paleobiology > National Museum of Natural History > The Smithsonian Institution [NHB, MRC 121] > P.O. Box 37012 > Washington DC 20013-7012 > > > (202) 633-1316 > slat...@si.edu > < > https://legacy.si.edu/owa/redir.aspx?C=HcyfurW5xkyXw2bWu-Td5xmRPtn7-c9I1lX > iVWbeJqzfhgybLjvFBckclIN237FXoGKsGXlPAGg.&URL=mailto%3aSlaterG%40si.edu> > www.eeb.ucla.edu/gslater > < > https://legacy.si.edu/owa/redir.aspx?C=HcyfurW5xkyXw2bWu-Td5xmRPtn7-c9I1lX > iVWbeJqzfhgybLjvFBckclIN237FXoGKsGXlPAGg.&URL=http%3a%2f%2fwww.eeb.ucla.edu > %2fgslater> >> > On 7/17/13 1:11 PM, "Liam J. Revell" wrote: > > >You can just change the bounds for optimization, see help(fitContinuous) > >for details. > > > >E.g.: > > > >fitContinuous(...,bounds=list(beta=c(0,100))) > > > >or something. Liam > > > >Liam J. Revell, A
Re: [R-sig-phylo] PGLS vs lm
"Maybe what we need here is an approach based on simultaneous equations (aka structural equation models), but I'm not aware whether this exists in a phylogenetic framework." Exactly! And it will need to incorporate "measurement error" in all variables as well as, eventually, uncertainly in the phylogenetic topology and branch lengths. Good luck, Ted From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Emmanuel Paradis [emmanuel.para...@ird.fr] Sent: Sunday, July 14, 2013 3:18 AM To: Joe Felsenstein Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] PGLS vs lm Hi all, I would like to react a bit on this issue. Probably one problem is that the distinction "correlation vs. regression" is not the same for independent data and for phylogenetic data. Consider the case of independent observations first. Suppose we are interested in the relationship y = b x + a, where x is an environmental variable, say latitude. We can get estimates of b and a by moving to 10 well-chosen locations, sampling 10 observations of y (they are independent) and analyse the 100 data points with OLS. Here we cannot say anything about the correlation between x and y because we controlled the distribution of x. In practice, even if x is not controlled, this approach is still valid as long as the observations are independent. With phylogenetic data, x is not controlled if it is measured "on the species" -- in other words it's an evolving trait (or intrinsic variable). x may be controlled if it is measured "outside the species" (extrinsic variable) such as latitude. So the case of using regression or correlation is not the same than above. Combining intrinsic and extinsic variables has generated a lot of debate in the literature. I don't think it's a problem of using a method and not another, but rather to use a method keeping in mind what it does (and its assumptions). Apparently, Hansen and Bartoszek consider a range of models including regression models where, by contrast to GLS, the evolution of the predictors is modelled explicitly. If we want to progress in our knowledge on how evolution works, I think we have to not limit ourselves to assess whether there is a relationship, but to test more complex models. The case presented by Tom is particularly relevant here (at least to me): testing whether group size affects brain size or the opposite (or both) is an important question. There's been also a lot of debate whether comparative data can answer this question. Maybe what we need here is an approach based on simultaneous equations (aka structural equation models), but I'm not aware whether this exists in a phylogenetic framework. The approach by Hansen and Bartoszek could be a step in this direction. Best, Emmanuel Le 13/07/2013 02:59, Joe Felsenstein a écrit : > > Tom Schoenemann asked me: > >> With respect to your crankiness, is this the paper by Hansen that you are >> referring to?: >> >> Bartoszek, K., Pienaar, J., Mostad, P., Andersson, S., & Hansen, T. F. >> (2012). A phylogenetic comparative method for studying multivariate >> adaptation. Journal of Theoretical Biology, 314(0), 204-215. >> >> I wrote Bartoszek to see if I could get his R code to try the method >> mentioned in there. If I can figure out how to apply it to my data, that >> will be great. I agree that it is clearly a mistake to assume one variable >> is responding evolutionarily only to the current value of the other >> (predictor variables). > > I'm glad to hear that *somebody* here thinks it is a mistake (because it > really is). I keep mentioning it here, and Hansen has published extensively > on it, but everyone keeps saying "Well, my friend used it, and he got tenure, > so it must be OK". > > The paper I saw was this one: > > Hansen, Thomas F & Bartoszek, Krzysztof (2012). Interpreting the evolutionary > regression: The interplay between observational and biological errors in > phylogenetic comparative studies. Systematic Biology 61 (3): 413-425. ISSN > 1063-5157. > > J.F. > > Joe Felsenstein j...@gs.washington.edu > Department of Genome Sciences and Department of Biology, > University of Washington, Box 355065, Seattle, WA 98195-5065 USA > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ > ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] PGLS vs lm
I think the issue is largely one of conceptualizing the problem. People often view body size as an "independent variable" when analyzing brain size, but obviously this is a serious oversimplificaiton -- usually done for statistical convenience -- that does not reflect the biology (yes, I have also done this!). Moreover, brain mass is part of body mass, so if you use body mass per se as an independent variable then you have potential part-whole correlation statistical issues. I would think carefully about what you are really wanting to do (e.g., regression vs. correlation vs. ANCOVA), and check over this paper: Ives, A. R., P. E. Midford, and T. Garland, Jr. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology 56:252-270. And maybe this one: Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265-292. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Tom Schoenemann [t...@indiana.edu] Sent: Thursday, July 11, 2013 11:19 AM To: Emmanuel Paradis Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] PGLS vs lm Thanks Emmanuel, OK, so this makes sense in terms of the math involved. However, from a practical, interpretive perspective, shouldn't I assume this to mean that we actually cannot say (from this data) whether VarA and VarB ARE actually associated with each other? In the real world, if VarA is causally related to VarB, then by definition they will be associated. Doesn't this type of situation - where the associations are judged to be statistically significant in one direction but not in the other - suggest that we actually DON'T have confidence that - independent of phylogeny - VarA is associated with VarB? Putting this in the context of the actual variables involved, doesn't this mean that we actually can't be sure brain size is associated with social group size (in this dataset) independent of phylogeny? I notice that the maximum likelihood lambda estimates are different (though I'm not sure they are significantly so). I understand this could mathematically be so, but I'm concerned with how to interpret this. In the real world, how could phylogenetic relatedness affect group size predicting brain size, more than brain size predicting group size? Isn't this a logical problem (for interpretation - not for the math)? In other words, in evolutionary history, shouldn't phylogeny affect the relationship between two variables in only one way, which would show up whichever way we approached the association? Again, I understand the math may allow it, I just don't understand how it could actually be true over evolutionary time. Thanks in advance for helping me understand this better, -Tom On Jul 11, 2013, at 5:12 AM, Emmanuel Paradis wrote: > Hi Tom, > > In an OLS regression, the residuals from both regressions (varA ~ varB and > varB ~ varA) are different but their distributions are (more or less) > symmetric. So, because the residuals are independent (ie, their covariance is > null), the residual standard error will be the same (or very close in > practice). > > In GLS, the residuals are not independent, so this difference in the > distribution of the residuals affects the estimation of the residual standard > errors (because we need to estimate the covaraince of the residuals), and > consequently the associated tests. > > Best, > Emmanuel > > Le 11/07/2013 11:03, Tom Schoenemann a �crit : >> Hi all, >> >> I ran a PGLS with two variables, call them VarA and VarB, using a >> phylogenetic tree and corPagel. When I try to predict VarA from VarB, I get >> a significant coefficient for VarB. However, if I invert this and try to >> predict VarB from VarA, I do NOT get a significant coefficient for VarA. >> Shouldn't these be both significant, or both insignificant (the actual >> outputs and calls are pasted below)? >> >> If I do a simple lm for these, I get the same significance level for the >> coefficients either way (i.e., lm(Var
Re: [R-sig-phylo] Phylosignal with phylo.d
Hi Miguel, I have not used this package. Is d the estimated parameter for an OU model? If so, then it would seem to be paramaterized differently from most implementations. In general, when one is using an OU model (d), Grafen's rho, Pagel's lambda, or an ACDC model (which we called g in Blomberg et al. 2003), a value of one (unity) gives the original tree, i.e., exactly the amount of signal (averaged across the tree) that you would expect under Brownian motion. A value of zero should return a star phylogeny (generally with contemporaneous tips, but I am not sure for all implementations of the above transforms). A value greater than one indicates more signal that expected under Brownian motion. You can get this in real data sets (see Blomberg et al. 2003). You can also get it in simulated data sets occasionally (rarely) by chance. Or, if you simulate under Brownian motion, and then manually add a constant to the phenotypes of all members of one or more entire clades, then you should be able to get the parameter estimate to exceed one. Note that the K statistic of Blomberg et al. (2003) behaves the same way. K = 0 indicates a star best fits the data (no phylogenetic signal), K = 1 indiucates the specified tree and a Brownian motion model fits the data well, and K > 1 indicates more signal than expected under Brownian motion. Again, one sometimes gets K values greater than one (highest we saw was 4.02). Blomberg, S. P., T. Garland, Jr., and A. R. Ives. 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57:717-745. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of miguel.ve...@uv.es [miguel.ve...@uv.es] Sent: Friday, May 17, 2013 8:56 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Phylosignal with phylo.d Dear list: I am using phylo.d in caper package and the Estimated D is lower than 0 (D=-0.71), suggesting a phylogenetic signal greater than Brownian expectation. However, the Probability of E(D) resulting from Brownian phylogenetic structure is p=0.98. Should I interpret this p-value as non-significant? or should I interpret it as a significant p=1-0.98=0.02? I am asking this question because I have never been able to detect a significant phylosignal higher than BM even with simulated data. Estimated D : -0.7107697 Probability of E(D) resulting from no (random) phylogenetic structure : 0 Probability of E(D) resulting from Brownian phylogenetic structure : 0.98 Thanks Miguel Verdu -- Centro de Investigaciones sobre Desertificacion (CSIC-UV-GV) Carretera Moncada - Náquera, Km. 4,5 Apartado Oficial 46113 Moncada (Valencia) Spain Tel +34 96 3424204 Fax +34 96 3424160 www.uv.es/verducam ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] pagel's lambda vs. abouheif's Cmean
Have you tried the randomization test of Blomberg et al. (2003), which should be in picante? Blomberg, S. P., T. Garland, Jr., and A. R. Ives. 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57:717-745. Also, the tests you describe are getting at somewhat different null hypotheses. We mention this here: Fusco, G., T. Garland, Jr., G. Hunt, and N. C. Hughes. 2011. Developmental trait evolution in trilobites. Evolution 66:314-329. Page 322: "Note that in all cases shown in Table 3, the likelihood of the star is actually higher than that for the specified phylogenetic tree, which would suggest that the traits do not exhibit phylogenetic signal. This apparent discrepancy versus the significant randomization tests for phylogenetic signal can be explained by at least two, not-mutually exclusive possibilities. First, comparing likelihoods is not equivalent to testing a null hypothesis of no phylogenetic signal. In particular, model fitting can be sensitive to local departures from the assumed model. For example, although many closely related pairs of species (i.e., those connected by short branch lengths) have similar protaspid growth increments (e.g., Amphilichas sp. [Ax] and Hemiarges turneri rasettii [Ht], Hyperbolochilus cf. marginauctum [Hm] and “Paraplethopeltis” n. sp. A [Px], Protopliomerella contracta [Pc] and Pseudocybele nasuta [Pn]; see Fig. 7A), one pair of closely related species, Arthricocephalus chauveaui (Ac) and Duyunaspis duyunensis (Dd), is strongly dissimilar. This one anomalous pair will strongly reduce the likelihood of the BM model but have less effect on randomization tests, which are expected to be less sensitive to outliers. Of these two species, A. chauveaui (Ac) has an unusually large value, as compared with other species nearby in the phylogenetic tree (Fig. 6). Deleting this one tip changes the results as expected: the likelihood of the star tree becomes lower than that of the hierarchical tree (see Table 3). Second, the level of signal, as indicated by the K statistic, is not large for any trait (compare with values in Blomberg et al. 2003)." Finally, other ways to test for signal besides lambda might perform better, such as an OU model. Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Ellis, Vincenzo A. (UMSL-Student) [vincenzoel...@mail.umsl.edu] Sent: Tuesday, April 23, 2013 10:38 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] pagel's lambda vs. abouheif's Cmean Hi everyone, In several datasets I have been looking at recently I generally find good agreement between Pagel's lambda and Abouheif's Cmean, i.e., if one says there is no phylogenetic signal so does the other. However, every so often a trait pops up and lambda and Cmean disagree, with lambda saying the trait is close to Brownian motion (phylogenetic signal present) and Cmean saying it's no different than random (no phylogenetic signal present). Does anyone know why that is happening? A recent paper in MEE by Munkenmuller et al. (2012) shows that lambda and Cmean correlate quite well so I'm at a bit of a loss. Thanks for any insights, Vincenzo Here's some code reproducing one of the contradictions I'm referring to: # Different results with Pagel's lambda and Abouheif's Cmean library(ape) tree <- read.tree(text="((Empidonax_virescens:64.912086,((Vireo_olivaceus:12.522261,Vireo_griseus:12.522261):34.717453,((Parus_carolinensis:15.623391,Baeolophus_bicolor:15.623391):28.479151,(Thryothorus_ludovicianus:43.329094,(((Icteria_virens:15.297336,Wilsonia_citrina:5.37719,(Dendroica_discolor:5.020348,Parula_americana:5.020348):0.356842):3.365272,(Mniotilta_varia:8.691948,(Vermivora_pinus:8.415811,Oporornis_formosus:8.41581):0.276138):0.050514):0.879735,Helmitheros_vermivorum:9.622197):0.673293,Seiurus_aurocapilla:10.29549):5.001846):0.986673,(Icterus_spurius:11.234447,Spizella_passerina:11.234447):5.049561):5.065457,((Cardinalis_cardinalis:17.182371,(Piranga_rubra:9.425619,Piranga_olivacea:9.425619):7.756752):1.264597,Passerina_cyanea:18.446968):2.902498):21.979624):0.773453):3.137173):17.672377):16.952915,Coccyzus_americanus:81.865005);") tree.pruned <- dr
Re: [R-sig-phylo] trait correlations with PICs
Exactly. That's why I wrote: "A problem is that then the statistical model is unclear/confusing." But for nuisance variables in a multuple-regression type of model, I see no problem putting them on a star phylogeny if they are not phylogenetically distrubuted (e.g., assuming that a particular technique for measuring, say, home range size, does not tend to be applied in a clade-specific [biased] fashion). So far as I am aware, Nobody knows exactly jhow to do this, nor has tried to implement this, with PGLS or phylogenetic regression models! We mention this in Garland and Ives (2000). Cheers, Ted From: Alejandro Gonzalez [alejandro.gonza...@ebd.csic.es] Sent: Wednesday, April 17, 2013 9:38 AM To: Theodore Garland Jr Cc: Anne Kempel; r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] trait correlations with PICs Can I risk a question here, maybe a result of ignorance... "With IC, it is easier to employ different sets of branch lengths for different traits (e.g. Garland et al., 1992; Lovegrove, 2003; Rezende et al., 2004), which may be particularly useful when one trait does not actually show phylogenetic signal (e.g. Tieleman et al., 2003; Rheindt et al., 2004) and/or for traits that are only nuisance variables, such as details of measurement or calculation methods that differ among studies (e.g. Wolf et al., 1998; Perry and Garland, 2002; Rezende et al., 2004)." >From an evolutionary point of view, what does it mean when using different >sets of branch lengths for different traits? If we're testing co-evolution of >traits, I would tend to think - and I may very well be wrong - that the >interest would be in estimating a single rate describing this co-evolution and >by employing different branch lengths would we not loose the opportunity to >estimtate that rate? Cheers Alejandro On 17, Apr 2013, at 6:24 PM, Theodore Garland Jr wrote: For correlations per se, please check this paper: Ives, A. R., P. E. Midford, and T. Garland, Jr. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology 56:252-270. And here comes my question: is it allowed to calculate only PICs based on my tree for those traits having a signal, and to calculate "pseudo-PICs" based on a star phylogeny (independence between species) for those traits having no signal? This would allow me to correct the "signal-traits" but leave the other ones more or less as they are. The benefit: I then can correlate all traits with each other. Obviously, it is possible to do this. A problem is that then the statistical model is unclear/confusing. I could not find any reference doing this, and I wonder whether this is legal? We have discussed this here: Garland, T., Jr., A. F. Bennett, and E. L. Rezende. 2005. Phylogenetic approaches in comparative physiology. Journal of Experimental Biology 208:3015-3035. Page 3032: "With IC, it is easier to employ different sets of branch lengths for different traits (e.g. Garland et al., 1992; Lovegrove, 2003; Rezende et al., 2004), which may be particularly useful when one trait does not actually show phylogenetic signal (e.g. Tieleman et al., 2003; Rheindt et al., 2004) and/or for traits that are only nuisance variables, such as details of measurement or calculation methods that differ among studies (e.g. Wolf et al., 1998; Perry and Garland, 2002; Rezende et al., 2004)." Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Anne Kempel [kem...@ips.unibe.ch] Sent: Wednesday, April 17, 2013 5:37 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] trait correlations with PICs Dear all, I am dealing with plant traits of different plant species (28 plant species), and aim to correlate all traits with each other. I have a tree of my species, and some of those traits do have a phylogenetic signal, but most don't. Since I do not know, which trait depends on whom, I favour to do correlations instead of regressions - that's why I would like to use Phylogenetic independent contrasts (PIC) instead of PGLS. I have the impression that it does not make sense to calculate PICs for traits that do not have a signal. And here comes my question: is it allowed to calculate only P
Re: [R-sig-phylo] trait correlations with PICs
For correlations per se, please check this paper: Ives, A. R., P. E. Midford, and T. Garland, Jr. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology 56:252-270. >And here comes my question: is it >allowed to calculate only PICs based on my tree for those traits having >a signal, and to calculate "pseudo-PICs" based on a star phylogeny >(independence between species) for those traits having no signal? This >would allow me to correct the "signal-traits" but leave the other ones >more or less as they are. The benefit: I then can correlate all traits >with each other. Obviously, it is possible to do this. A problem is that then the statistical model is unclear/confusing. >I could not find any reference doing this, and I wonder >whether this is legal? We have discussed this here: Garland, T., Jr., A. F. Bennett, and E. L. Rezende. 2005. Phylogenetic approaches in comparative physiology. Journal of Experimental Biology 208:3015-3035. Page 3032: "With IC, it is easier to employ different sets of branch lengths for different traits (e.g. Garland et al., 1992; Lovegrove, 2003; Rezende et al., 2004), which may be particularly useful when one trait does not actually show phylogenetic signal (e.g. Tieleman et al., 2003; Rheindt et al., 2004) and/or for traits that are only nuisance variables, such as details of measurement or calculation methods that differ among studies (e.g. Wolf et al., 1998; Perry and Garland, 2002; Rezende et al., 2004)." Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Anne Kempel [kem...@ips.unibe.ch] Sent: Wednesday, April 17, 2013 5:37 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] trait correlations with PICs Dear all, I am dealing with plant traits of different plant species (28 plant species), and aim to correlate all traits with each other. I have a tree of my species, and some of those traits do have a phylogenetic signal, but most don't. Since I do not know, which trait depends on whom, I favour to do correlations instead of regressions - that's why I would like to use Phylogenetic independent contrasts (PIC) instead of PGLS. I have the impression that it does not make sense to calculate PICs for traits that do not have a signal. And here comes my question: is it allowed to calculate only PICs based on my tree for those traits having a signal, and to calculate "pseudo-PICs" based on a star phylogeny (independence between species) for those traits having no signal? This would allow me to correct the "signal-traits" but leave the other ones more or less as they are. The benefit: I then can correlate all traits with each other. I could not find any reference doing this, and I wonder whether this is legal? I am grateful for any comment and suggestion! Thanks a lot in advance and best wishes from Switzerland, Anne -- Dr. Anne Kempel Institute of Plant Sciences Altenbergrain 21 3013 Bern Switzerland ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] trait correlations with PICs
Hi all, I will attempt some clarification. In the context of phylogenetic regression models, the three most common types of models are: 1. Ordinary Least Squares (OLS) = conventional statistics = assumes a star phylogeny with contemporaneous tips. 2. Phylogenetic Generalized Least Squares (PGLS) = GLS where you use a phylogenetic tree to compute the expected variance-covariance matrix of residuals. First developed by Grafen (1989) (using GLIM, an uncommon stats package). Later Martins and Hansen (1997), Garland and Ives (2000), others, Lavin et al. (2008): Grafen, A. 1989. The phylogenetic regression. Phil. Trans. Royal. Soc. Lond. B 326:119-157. Garland, T., Jr., and A. R. Ives. 2000. Using the past to predict the present: Confidence intervals for regression equations in phylogenetic comparative methods. American Naturalist 155:346-364. Lavin, S. R., W. H. Karasov, A. R. Ives, K. M. Middleton, and T. Garland, Jr. 2008. Morphometrics of the avian small intestine, compared with non-flying mammals: A phylogenetic perspective. Physiological and Biochemical Zoology 81:526-550. [provides Matlab Regressionv2.m, released as part of the PHYSIG package] Grafen called this the "standard regresison," which confused people (self included!). What he called the "phylogenetic regression" used a "transformation parameter" that he named rho and was purely statistical, not tied to a model of evolution. This method is not simply PGLS (see below), although people keep calling it that. We have a discussion of this in the Appendix to Lavin et al. (2008). It also gives the history of all these methods. As we say in the paper: "Because the foregoing analyses assume either no (OLS) or relatively strong (PGLS) phylogenetic signal, we also performed an analysis in which the strength of phylogenetic signal in the residual variation was estimated simultaneously with the regression coefficients (e.g., see Grafen 1989; Freckleton et al. 2002; Chown et al. 2007; Duncan et al. 2007)." The additional parameter that is estimated is Grafen's rho, Pagel's lambda, or what we call "d" when it is tied to an OU model (first suggested by Felsenstein 1988 and used in simulations by Garland et al. 1993) or "g" when it is tied to an ACDC model of residual character evoluton (see Blomberg et al. 2003). 3. Phylogenetic Regression with a ??? Transformation/Model assumed. This all started with Grafen (1989). He also dealt with soft polytomies (lack of knowledge about branching order), and this made his paper even more sophisticated but complicated and a bit hard to understand at the time (it was ahead of its time!). In Lavin et al. (2008) we write (page 544): "Deriving a sensible terminology to discuss estimation under these different transform models requires making a distinction between statistical models and estimation techniques used to fit models to data. All of these transform models are statistical models in that they describe a statistical distribution and its parameters. In contrast, OLS and GLS are estimation techniques. Several authors refer to one or more of the transform models as GLS models (or PGLS models, for phylogenetic GLS models), although GLS is an estimation procedure. This is particularly confusing because GLS actually cannot be used to estimate the parameters of these transform models because they all contain parameters in the variance-covariance matrix that must be estimated. To reduce confusion, our preference is to refer to transform models as distinct from the estimation approaches that can be applied to them. We will break this convention only when using the established monikers “OLS model” and “GLS model” (i.e., the Brownian motion model) because in these cases there is a one-to-one match between the structure of the model and the appropriate estimation technique." The Matlab Regressionv2.m program we provide uses REML to estimate parameters in the four transform models: Grafen’s rho, Pagel’s lambda, OU, and ACDC (OU and ACDC are the same if you have contempoiraneous tips). Note that the version of the OU model implemented in various programs may differ a bit! See also Butler and King work, O'Meara, and probably others! Enough rambling! Hope that helps!!! Cheers, Ted Theodore Garland, Jr., Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Skype: theodoregarland Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Inquiry-based Middle School Lesson Plan: "Born to Run: Artificial Selection Lab" http://www.indiana.edu/~ensiweb/lessons/BornToRun.html From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Alejandro Gonzalez
Re: [R-sig-phylo] Ancestral state estimates of continuous traits
Yes, we have always used REML for this! Cheers, Ted From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Emmanuel Paradis [emmanuel.para...@ird.fr] Sent: Friday, March 15, 2013 11:21 PM To: Liam J. Revell; r-sig-phylo-boun...@r-project.org; Alejandro Gonzalez Cc: R-phylo Mailing-list Subject: Re: [R-sig-phylo] Ancestral state estimates of continuous traits Alejandro and Liam, You can also try method="REML" which gives much better estimates of sigma^2 than ML. I think I'll make it the default for continuous characters. Best, Emmanuel -Original Message- From: "Liam J. Revell" Sender: r-sig-phylo-boun...@r-project.org Date: Fri, 15 Mar 2013 14:23:20 To: Alejandro Gonzalez Cc: R-phylo Mailing-list Subject: Re: [R-sig-phylo] Ancestral state estimates of continuous traits Hi Alejandro. That must be a bug in ace(...,method="GLS"). I would also suggest you check out fastAnc in the phytools package. It uses ace(...,method="pic") internally to take advantage of the fact that the contrasts estimate at the root and the MLE assuming Brownian evolution are the same. It re-roots the tree at all internal nodes which sounds computationally intensive but is actually much faster than getting the MLEs via numerical optimization. fastAnc also uses equation (6) from Rohlf (2001) to get the correct 95% interval on the estimates. My analysis (http://blog.phytools.org/2013/02/new-version-of-fastanc-new-build-of.html) suggests that the 95% CIs from ace(...,method="ML") are too small. (This is probably because they rely on asymptotic properties of likelihood that are not satisfied for the relatively small size of most phylogenetic datasets.) All the best, Liam Liam J. Revell, Assistant Professor of Biology University of Massachusetts Boston web: http://faculty.umb.edu/liam.revell/ email: liam.rev...@umb.edu blog: http://blog.phytools.org On 3/15/2013 8:18 AM, Alejandro Gonzalez wrote: > Hello, > > I am using ape to obtain ancestral state estimates for continuous > traits. Two options are available, either maximum likelihood or a GLS > method. I am comparing the results of both methods and one difference > between the two puzzles me, I hope someone can enlighten me. Under GLS > the ancestral state estimate at the root has virtually identical values > for the point estimate and 95% confidence intervals, here is one > example: 100 1.7848301 1.78483005 1.7848301 (100 is the root node > number, then the ancestral state estimate and 95% CIs, respectively). > However, when I use maximum likelihood for the same ancestral sate > estimate, with the same data and tree I get the following result for the > root : 100 1.6570119 0.86612615 2.4478977 (numbers in the same order > as above). > Any ideas as to why GLS gives such narrow confidence intervals for the > ancestral state estimate at the basal node? > > Cheers > > Alejandro > __ > > Alejandro Gonzalez Voyer > > Post-doc > > Estaci�n Biol�gica de Do�ana > Consejo Superior de Investigaciones Cient�ficas (CSIC) > Av Am�rico Vespucio s/n > 41092 Sevilla > Spain > > Tel: + 34 - 954 466700, ext 1749 > > E-mail: alejandro.gonza...@ebd.csic.es > > Web site (Under construction): > > Personal page: http://consevol.org/members/alejandro_combo.html > > Group page: http://consevol.org/people.html > > For PDF copies of papers see: > > http://csic.academia.edu/AlejandroGonzalezVoyer > > > > [[alternative HTML version deleted]] > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] Some questions about pPCA - phylogenetic anti-signal
You can test for anti-signal using the randomization procedures of:
Blomberg, S. P., T. Garland, Jr., and A. R. Ives. 2003. Testing for
phylogenetic signal in comparative data: behavioral traits are more labile.
Evolution 57:717-745.
See page 719.
Cheers,
Ted
Theodore Garland, Jr., Professor
Department of Biology
University of California, Riverside
Riverside, CA 92521
Office Phone: (951) 827-3524
Facsimile: (951) 827-4286 = Dept. office (not confidential)
Email: tgarl...@ucr.edu
http://www.biology.ucr.edu/people/faculty/Garland.html
http://scholar.google.com/citations?hl=en&user=iSSbrhwJ
Inquiry-based Middle School Lesson Plan:
"Born to Run: Artificial Selection Lab"
http://www.indiana.edu/~ensiweb/lessons/BornToRun.html
From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on
behalf of Francois KECK [francois.k...@thonon.inra.fr]
Sent: Monday, February 11, 2013 5:47 AM
To: r-sig-phylo@r-project.org
Subject: Re: [R-sig-phylo] Some questions about pPCA
Hello Thibaut,
Thank you for these clarifications.
About 2: I understand how to use the abouheif test to detect the
phylogenetic signal (up to now I used it with the patristic matrix of
proximities). But I don't know how to use it to test the anti-signal.
Absence of signal is not anti-signal, or missed something?
Cheers
François
> Hello François,
>
> 1. In pPCA, the sum of the eigenvalues is often meaningless, because it can
> be a mixture of large positive and negative values. So this ratio is no
> longer relevant. Selection of eigenvalues can be based on the amount of
> variance and autocorrelation (Moran's I) represented (each eigenvalue is a
> product of the two). See summary.ppca and screeplot.ppca.
>
> 2. The best way is testing positive/negative phylogenetic autocorrelation
> ("global/local" structures in the paper's terminology) beforehand. See
> section 3.1 of the vignette "Quantifying and testing phylogenetic signal" -
> abouheif.moran will test all variables at once (just make sure to use the
> same measure of proximity in the pPCA).
>
> 3. As you suspected, testing phylogenetic signal of pPCA components is
> meaningless, as these synthetic variables are already optimized for
> phylogenetic signal. Estimating ancestral states is always possible; I can
> see at least two ways of doing it: a) reconstruct the ancestral state of
> every traits, and then compute the coordinates of the nodes on the pPCA axis
> using the loadings of the analysis. In this case, nodes are used as
> 'supplementary individuals'. b) reconstruct directly the principal component
> of pPCA; in this case, the component needs to have a clear-cut interpretation.
>
> Cheers
>
> Thibaut
>
>
> From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org]
> on behalf of Francois KECK [francois.k...@thonon.inra.fr]
> Sent: 11 February 2013 10:05
> To: r-sig-phylo@r-project.org
> Subject: [R-sig-phylo] Some questions about pPCA
>
> Dear all,
> I'm a new subscriber to this list since I just started to play with
> phylogenetic data with R. The task is facilitated by reading the
> excellent book of E. Paradis. However I recently discovered the pPCA
> method (as introduced by Jombart et al. 2010) and i'm very interested in
> it to work on phylogenetic signal but I still have some questions...
>
> 1. I'm a long time user of ade4 to perform multivariate analysis. For a
> classic PCA I usually calculate the % of variance taking account by each
> axis using :
> myPCA$eig/sum(myPCA$eig) * 100
> I'd just like to be sure I can do the same with a pPCA, using absolute
> values of the eigenvalues, e.g.:
> abs(myPPCA$eig)/sum(abs(myPPCA$eig)) * 10
>
> 2. In their paper, Jombart et al. present some figures where they
> sometimes exclude directly the local or the global principal component
> because they know it doesn't exist (these are simulated data). Is there
> a way to test the global vs the local component with "real data"? With
> my own data I have a very low "local eigenvector" so I wonder if I could
> only focus on global structure. Can I justify this choice with statistics?
>
> 3. I think it could be interesting to play with the species coordinates
> especially with the global component. But does it make sense to assess
> the phylogenetic signal or to estimate ancestral characters on these
> constrained data? I'm a little doubtful about that and your point of
> view is welcome.
>
> Thank you for your help.
>
> François KECK
>
> ___
> R-sig-phylo mailing list - R-sig-phylo@r-project
Re: [R-sig-phylo] R: Re: From ClustalW2 Tree to Heat Map in R
Hi Pas, K is just a descriptive statistic for tip data on a tree with some specified branch lengths. Obviously, if you change the branch lengths (or the topology or the tip data), then K will be different. How much different depends on the details. It is likely that certain branch-length transformations (whether purely statistical or designed to mimic something like an OU process) will tend to make K either larger or smaller, at least when checked cross a large number of examples. Somebody would need to do a lot of simulations to sort this out. Some related relevant things are here: Revell, L. J., L. J. Harmon, and D. C. Collar. 2008. Phylogenetic signal, evolutionary process, and rate. Syst. Biol. 57:591-601. In the original Blomberg et al. (2003) paper we avoided this issue entirely by only reporting K using whatever tree the original paper used. We wanted to be able to compare K values among traits and studies in some sort of simple and "fair" way. If lambda and OU transforms are leading to very different K values (how different?) that does not mean one K value is better than the other. But it does mean you would want to be careful if you tried to compare your K value with other studies that did not use lambda and/or OU transforms. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of pasquale.r...@libero.it [pasquale.r...@libero.it] Sent: Thursday, January 24, 2013 12:14 PM To: r-sig-phylo@r-project.org Mailing-list Subject: [R-sig-phylo] R: Re: From ClustalW2 Tree to Heat Map in R Hi Gents and Lads, I have a very rapid question with a perhaps not-so-obvious reply. I'm in the process of testing a number of evolutionary models and estimating phylogenetic signal on a certain univariate data set. So something very basic and very simple. The point is, I found BM performs poorly as compared to OU (single peak) and lambda. Thus, I transformed the tree by using the fitted lamdba and/or alpha before calculating Blomberg et al's K statistic. Does this make sense? If yes, the competing models have similar Akaike weigths (OU = 0.5, lambda = 0.3) but give very different estimates of K when their fitted parameters are used to transform the tree branch lenghts. How does discriminate which K estimate is best? Translating in R-esque: require(geiger) require(phytools) fitContinuous(tree, x, model="OU") ## gives relatively low alpha (.07) phylosig(ouTree(tree,alpha=alpha),x,method ="K", test =T, nsim =1000) ## gives very low K fitContinuous(tree, x, model="lambda") ## gives very high lambda (.95) phylosig(lambdaTree(tree,lambda=lambda),x,method ="K", test =T, nsim =1000) ## gives very high K thanks for help, Pas ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] interpreting phylogenetic signal
The K value tells you how much signal you have, relative to a Brownian motion model of evolution. Please see the explanation and large compilation of values for different traits in Blomberg et al. (2003), Evolution (available on my webpage) For body size, K values on average did not differ statistically from 1.00. However, K is for continuous-valued traits. It does not make sense for many categorical traits! What, exactly, is you trait? cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Gabriel Yedid [gyedi...@gmail.com] Sent: Saturday, December 08, 2012 9:04 AM To: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] interpreting phylogenetic signal On Sat, Dec 8, 2012 at 9:36 PM, Steven Kembel wrote: > Hi, > If you want to test the hypothesis that there is non-random amount of > phylogenetic signal, you want the P-value (PIC.variance.P). > The output of phylosignal is described in the help for the function (type > ?phylosignal to view the help). > -Steve Yes, but is the p-value really all that matters? What about the magnitude of the K-values, or the ratio of the observed to random PIC variances, or the actual Z-score? I ran a similar analysis very recently (with multiPhylosignal) and got answers that said all the traits I was looking at had significant signal (p < 0.001), but the K-values were all less than one. However, my phylogenies are very large (>1000 tips) and the traits I'm looking at are discrete, so I don't know how that could be affecting this test. cheers, Gabe > Le 2012-12-08 à 00:40, E Pearson a écrit : > >> I calculated Blomberg's K using 'Kcalc' in 'picante' for a continuous trait >> and a phylogeny of 65 species. >> >> >> The K value is 1.079 and when I do the randomization test I get the >> following results. I'm a bit confused by such a large PIC.variance.rnd.mean >> = 8.031. >> >> >> Should I be drawing conclusions by P = 0.0009 instead? Rejecting the NULL >> of no phylogenetic signal? >> >> >> # K PIC.variance.obs PIC.variance.rnd.mean PIC.variance.P PIC.variance.Z >> >> # 1.079533 1.274221 8.0318440.000999001 >> -3.351957 >> >> [[alternative HTML version deleted]] >> >> ___ >> R-sig-phylo mailing list - R-sig-phylo@r-project.org >> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo >> Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ > > -- > Steven W. Kembel > Professeur régulier > Département des sciences biologiques > Université du Québec à Montréal > kembel.steve...@uqam.ca > +1 (514) 987-3000 poste 5855 > http://www.phylodiversity.net/skembel/ > > ___ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/ ___ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/
Re: [R-sig-phylo] variation in rates over time, unexpected message when using Brownie.lite
That number of 20 comes up from the simulations provided by both of these papers: Blomberg, S. P., T. Garland, Jr., and A. R. Ives. 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57:717-745. Freckleton, R. P., P. H. Harvey, and M. Pagel. 2002. Phylogenetic analysis and comparative data: a test and review of evidence. Am. Nat. 160:712-726. Note, however, that both of those sets of simulations assumed no measurement error in the phenotypic trait(s) and no error in the phylogenetic topology or branch lengths. Hence, they may well have underestimated how many species you need in the real world! As they say, your mileage may vary ... Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Brian O'Meara [bome...@utk.edu] Sent: Tuesday, September 18, 2012 1:25 PM To: Agus Camacho Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] variation in rates over time, unexpected message when using Brownie.lite That's a fairly small tree to be doing this with -- an old rule of thumb (I think from Pagel) is 20 tips per parameter. Common issues are zero length (or effectively zero length) branch lengths (what is min(phy$edge.length)?) and polytomies (methodologically not a problem, but many packages don't deal well with these). Brian ___ Brian O'Meara Assistant Professor Dept. of Ecology & Evolutionary Biology U. of Tennessee, Knoxville http://www.brianomeara.info Students wanted: Applications due Dec. 15, annually Postdoc collaborators wanted: Check NIMBioS' website Calendar: http://www.brianomeara.info/calendars/omeara On Tue, Sep 18, 2012 at 4:08 PM, Agus Camacho wrote: > Dear all, I tried to test something similar to Jason, but relating the > change in rates to the acquisition of a trait, instead to a specific date. > > For that, I used a phylogeny with ten taxa, without outgroup. > I used exactly the same script gently proposed by Liam but got the > following error message: > "Error in solve.default(model1$hessian) :Lapack routine dgesv: system is > exactly singular" > > I've been looking through the internet but was not able to identify what > causes this error. Any hint about that? > > BTW, anybody knows how to directly answer a message found in the digest of > the mail list, without leaving the thread? > > Below, I pasted the complete thread dealing with this topic. > > Cheers, > Agus > > Message: 9 > Date: Tue, 18 Sep 2012 00:13:00 -0400 > From: "Brian O'Meara" > To: Jason S > Cc: "R-sig-phylo@r-project.org" > Subject: Re: [R-sig-phylo] variation in rates over time > Message-ID: > < > cakywhkq-jjroqzdvfutphi4xopxam+aaacu64nldb3inxoz...@mail.gmail.com> > Content-Type: text/plain > > I agree with the suggestions so far. I just wanted to point out a few more > alternatives: > > You could use the geiger package to estimate the best scaling for the > tworatetree transformation to do this (should be equivalent to the earlier > solutions, though it would require running optimization). > > You could also use the OUwie package with a tree that has been given simmap > mappings using phytools. The advantage of this is that you could evaluate > Brownian models but you could also look at various Ornstein-Uhlenbeck > models (though note that while there is information about different > Brownian rates before and after a time slice, info about different alphas > (strength of pull parameter) and thetas (the attractor, aka "optimal > value") is rapidly (but not immediately) lost under an OU process). > > For completeness, especially for citations, note that O'Meara et al. (2006) > and Thomas et al. (2006) independently arrived at essentially the same > method, so it is worth reading both papers. > > Best, > Brian > > ___ > Brian O'Meara > Assistant Professor > Dept. of Ecology & Evolutionary Biology > U. of Tennessee, Knoxville > http://www.brianomeara.info
Re: [R-sig-phylo] alternative to Bartlett test of homogeneity of variances
Several. See these papers for starters: Garland, T., Jr. 1992. Rate tests for phenotypic evolution using phylogenetically independent contrasts. Am. Nat. 140:509-519. Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265-292. Hutcheon, J. M., and T. Garland, Jr. 2004. Are megabats big? Journal of Mammalian Evolution 11:257-276. Collar, D. C., P. C. Wainwright, T. J. Near. 2005. Comparative analysis of morphological diversity: does morphological disparity evolve at the same rate in two lineages of centrarchid fishes. Evolution 58:1783-1794. [first use of method by O'Meara et al., 2006] O'Meara, B. C., C. Ane, M. J. Sanderson, and P. C. Wainwright. 2006. Testing for different rates of continuous trait evolution using likelihood. Evolution 60:922-933. [full explanation of method used by Collar et al., 2005] Rezende, E. L., and J. A. F. Diniz-Filho. 2012. Phylogenetic analyses: comparing species to infer adaptations and physiological mechanisms. Comprehensive Physiology 2:639-674. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Jason S [jas2...@yahoo.com] Sent: Monday, September 17, 2012 1:08 PM To: R-sig-phylo@r-project.org Subject: [R-sig-phylo] alternative to Bartlett test of homogeneity of variances Hi there, Is there a PCM that is analogous to the Bartlett test of homogeneity of variances? Thanks, Jason Mustakas [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Comparative analyse and heritability
I am still a little unclear as to how to think about this trait, but it is an interesting problem that has also been discussed here: Garland, T., Jr., P. H. Harvey, and A. R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41:18-32. Pages 29-30" WHAT KINDS OF TRAITS CAN BE ANALYZED? In addition to the usual phenotypic traits (e.g., body size, metabolic rate), cultural (see Cavalli-Sforza and Feldman, 1981), environmental (e.g., soil or water pH, mean annual temperature), and other traits that are difficult to categorize (e.g., home range area) can be studied as long as they are passed on from ancestral to descendent species (or populations) and have a continuous distribution. For example, many environmental properties, such as latitude or mean annual rainfall, are not inherited in the conventional (genetic) sense. Nevertheless, they are inherited in the sense that organisms are born into environmental conditions and locations experienced by their parents at the time of birth. Thus, the ancestor of two species living in a desert may also have lived in a desert (cf. Huey, 1987), or the ancestor of one high-latitude and one equatorial species may have lived at midlatitude. Similarly, if an environmental characteristic is determined solely (without externally imposed constraints) through a process of habitat selection, and if species differences in habitat selection are genetically based, then species differences in the environmental trait will be genetically based as well. Alternatively, if variation in some (genetically based) phenotypic trait can be used as a precise indicator of some environmental characteristic, then that phenotypic trait may be used as a surrogate for the environmental characteristic. For example, toe fringes in lizards might indicate occupancy of sandy habitats. Unfortunately, this is not unfailingly the case; some species that glide through the air or that run across water also possess toe fringes (Luke, 1986). Finally, paleoclimatological and historical biogeographical data might be used in conjunction to indicate environmental characteristics of hypothetical ancestral (as opposed to tip) species, but this takes us into the realm of other comparative methods, such as those based on minimum evolution reconstructions of ancestors (Huey, 1987; Harvey and Pagel, 1991; Maddison, 1991; Martins and Garland, 1991). In any case, techniques for correlating phenotypes with environmental characteristics require further study. And, in these papers, we used a hierarchical tree to compute phylogenetically independent contrasts for one trait and a collapsed tree to make a star for another, exactly as you are suggesting: Bonine, K. E., and T. Garland, Jr. 1999. Sprint performance of phrynosomatid lizards, measured on a high-speed treadmill, correlates with hindlimb length. Journal of Zoology, London 248:255-265. Perry, G., and T. Garland, Jr. 2002. Lizard home ranges revisited: effects of sex, body size, diet, habitat, and phylogeny. Ecology 83:1870-1885. Wolf, C. M., T. Garland, Jr., and B. Griffith. 1998. Predictors of avian and mammalian translocation success: reanalysis with phylogenetically independent contrasts. Biological Conservation 86:243-255. The trick is to first put that other trait on the same tree as you use for "regular" traits and then collapse it so all of the internode branches have zero length (and the tips are contemporaneous). That way (perhaps depending on exactly how the program you use is keeping track of things), the contrasts will all be isomorphic (same tips or nodes being contrasted) for both traits. Then you can do a correlation or regression (always through the origin) in which one or more traits is on a hierarchical tree and one or more is on a star phylogeny. We did this for non-phylogenetic nuisance variables (such as calculation method for home range size), but you could also do it for a "real" trait that was clearly non-phylgenetic (if there is such a thing -- see Joe's comment). Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Joe Felsenstein [j...@gs.washington.edu] Sent: Thursday, September 06, 2012 2:38 AM To: tristan.lefeb...@gmail.com Cc: r-sig-phylo@r-pro
Re: [R-sig-phylo] testing binomial characters
Check this: Ives, A. R., and T. Garland, Jr. 2010. Phylogenetic logistic regression for binary dependent variables. Systematic Biology 59:9-26. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Jordan Golubov [gfjor...@correo.xoc.uam.mx] Sent: Wednesday, August 29, 2012 12:23 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] testing binomial characters Dear All, I am trying to test whether a binomial character state (photoblastic seeds vs indifferent responsive seeds) in species of Cactaceae is affected by seed mass or length. 1) Is there a meas of identifying phylogenetic signal for a two character state? 2) Is there a way of testing what factor (both continuous seed mass/length) affects photoblastic rersponse, maybe a GLMM? Thanks -- Dr. Jordan Golubov Lab. Ecologia, Sistematica y FisiologiaVegetal Departamento El Hombre y Su Ambiente Universidad Autonoma Metropolitana Xochimilco Calz. del Hueso 1100, Col. Villa Quietud, Coyoacán 04960, México D. F. México "It is the mark of an educated mind to be able to entertain a thought without accepting it". >Aristotle ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] pPCA - global and local components
That seems like it would be OK, at least if you think it is OK for nonphylogenetic PCA. An alternative is to simulate data along your phylogeny, analyze it the same way, do it a couple thousand times, then make an empirical null distribution of, say, the eigenvalues when the data have no correlation on average but increased variance in the values of correlations caused by the phylogenetic hierarchy. This is discussed in our very old PHYLOGR package. However, you will need to make some decisions about the "branch lengths" to use for your individuals within species, represented by a bunch of mini-star phylogenies. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html http://scholar.google.com/citations?hl=en&user=iSSbrhwJ Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Franck Stefani [fopstef...@gmail.com] Sent: Wednesday, August 22, 2012 3:50 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] pPCA - global and local components Hi, Among the graphical outputs of the pPCA, there is the scree plot showing the global and local components. I would like to know what are the criteria to define the number of GPC or LPC to interpret ? Can we use a broken stick model? Cheers, Franck [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Can PGLS cope with collinearity between explanatory variables?
The issue of collinearity of independent variables is neither better nor worse with PGLS as opposed to OLS. Statistical significance per se of a correlation between X variables is not really the issue. How strong is the correlation? Most sources suggest that it needs to be greater than 0.7-0.8 in magnitude to cause serious problems. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Xu Han [duck_han365...@hotmail.com] Sent: Friday, August 17, 2012 12:33 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Can PGLS cope with collinearity between explanatory variables? Hi all, I am testing a correlation between two explanatory variables and a response variable using PGLS. All of the variables are continuous. My model is Log female body size ~ Log egg size * Log clutch size. However, there is a significant negative correlation between egg size and clutch size. Can PGLS cope with collinearity between explanatory variables? Is there any way that I can apply something like principal component analysis to PGLS models? Thanks, Xu [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] PIC or PGLS for genome-wide SNP screening
>Is this the way in which one decides for OLS vs. PGLS? If you have the same set of independent variables, then you just prefer the one (OLS or PGLS) with the higher likelihood. So far as I am told by Joe Felsenstein, you cannot do a ln maximum likelihood ratio test because the number of parameters is the same, although this paper seems to suggest otherwise: Mooers, A. O., S. M. Vamosi, and D. Schluter. 1999. Using phylogenies to test macroevolutionary hypotheses of trait evolution in cranes (Gruinae). American Naturalist 154:249-259. For two models with the same set of independent variables, AIC does not add anything for you. If you go to something like Regression with an OU process modeled for the residuals, then you do have an additional parameter being estimated and so you can do an ln maximum likelihood ratio test of that model versus OLS and versus PGLS. For example, see: Lavin, S. R., W. H. Karasov, A. R. Ives, K. M. Middleton, and T. Garland, Jr. 2008. Morphometrics of the avian small intestine, compared with non-flying mammals: A phylogenetic perspective. Physiological and Biochemical Zoology 81:526-550. [provides Matlab Regressionv2.m, released as part of the PHYSIG package] Gartner, G. E. A., J. W. Hicks, P. R. Manzani, D. V. Andrade, A. S. Abe, T. Wang, S. M. Secor, and T. Garland, Jr. 2010. Phylogeny, ecology, and heart position in snakes. Physiological and Biochemical Zoology 83:43-54. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Mattia Prosperi [ahn...@gmail.com] Sent: Wednesday, May 23, 2012 8:05 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] PIC or PGLS for genome-wide SNP screening Dear all, I am working on a data set composed of bacterial genomic sequences (a few genes) associated to phenotypic values (in-vitro resistance to antibiotics, a numerical value discretised into a binary class). Of note, the bacterial isolates were sampled non-uniformly at different times and locations, thus with a possible sampling bias. The data set is ~1,000 variables and ~1,000 observations. I have been applying several methods for developing a model to predict antibiotic resistance from the single nucleotide polymorphisms (SNP) extracted from a multiple alignment, applying classical statistical learning and feature selection methods. Eventually, I found that a logistic regression with main effects, where the variables were selected first by a univariable chi-square screening and then by AIC stepwise, was as good as other more complex and non-linear methods (such as random forests) by comparing different loss function (AUROC, specificity, sensitivity) distributions upon multiple cross-validation runs. Also, the SNP sets selected by different approaches were highly similar and consistent across several bootstrap evaluations. I found that a few relevant (even after Bonferroni correction) SNP were located in gene regions that are not supposed to be related with antibiotic resistance. I thought that this might be a consequence of neutral mutations that became fixed in the population by chance after a genetic bottleneck (e.g. antibiotic pressure). I'd like to understand if such SNP that is associated to antibiotic resistance (and actually not expected to be) is indeed just a random mutation of an early isolate that was carrying the true resistance SNP (in another gene region) and that was selected by the antibiotic pressure, thus transfering both the true resistance SNP and the "hitchhicking" ones to the offspring. Unfortunately it is not easy to cross-tabulate SNP in different genes because not all isolates have been sequenced the same set of genes. In order to check for fake/true SNP associated to resistance, I thought I might use a PIC or PGLS approach (after estimating a phylogenetic tree from the multiple alignment), in the same settings as the original analysis, i.e. a model selection approach with both feature and performance evaluation (well, since the coefficients of PGLS/OLS are the same, it's just a matter of standard errors and feature set selection), regressing the resistance class as a dependent variable and using the SNP as covariates. Is this a reasonable approach? Does it make sense to set up -for in
Re: [R-sig-phylo] Best way to test correlation between discrete and continuous variables ?
Dear Julien, I'll just add two other papers to your reading list (both available at my web page): Garland, T., Jr., A. W. Dickerman, C. M. Janis, and J. A. Jones. 1993. Phylogenetic analysis of covariance by computer simulation. Systematic Biology 42:265-292. And the Appendix of this paper: Lavin, S. R., W. H. Karasov, A. R. Ives, K. M. Middleton, and T. Garland, Jr. 2008. Morphometrics of the avian small intestine, compared with non-flying mammals: A phylogenetic perspective. Physiological and Biochemical Zoology 81:526-550. [provides Matlab Regressionv2.m, released as part of the PHYSIG package] Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Marguerite Butler [mbutler...@gmail.com] Sent: Wednesday, May 16, 2012 12:27 PM To: Julien Lorion Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Best way to test correlation between discrete and continuous variables ? Dear Julien, There is no problem with applying an ANOVA within a phylogenetic framework. This is essentially phylogenetic GLS, which you can implement easily with APE. You can have a look at Emmanuel's book (which was just recently came out in the second edition Nov. 2011, by the way). http://www.amazon.com/Analysis-Phylogenetics-Evolution-Emmanuel-Paradis/dp/1461417422/ref=sr_ob_11?s=books&ie=UTF8&qid=1337196146&sr=1-11 In essence, you are looking at the phylogeny as a source of "correlated errors" which you are "correcting for" under some assumed model of evolution -- either Brownian motion or Ornstein Ulenbeck. It is viewed as noise which is controlled for in order to see the pattern from ecology, etc. The mechanics of how to incorporate the phylogenetic covariance matrix into the linear model is explained in the appendix of my paper: Butler M.A. Schoener T.W., and Losos J.B. (2000) The relationship between habitat type and sexual size dimorphism in Greater Antillean Anolis lizards. Evolution 54(1):259-272. DOI: http://dx.doi.org/10.1554/0014-3820(2000)054[0259:TRBSSD]2.0.CO;2 Another approach to analyzing the same kind of data is to view the evolution of the quantitative character as being influenced by a number of factors (for example, habitat, symbionts, etc.), which can be thought of as "selective regimes" which influence the evolution of body size. You can then create explicit biological hypotheses which are translated to mathematical models, and test these hypotheses against each other for the best explanation of the data. This approach has software package developed for it called "OUCH" which is available in R. It is explained and illustrated in this paper: Butler M.A. and King A.A. (2004) Phylogenetic comparative analysis: a modeling approach for adaptive evolution. The American Naturalist 164(6):683-695. DOI: 10.1086/426002 Appologies for the shameless self-promotion:). Marguerite On May 15, 2012, at 9:53 PM, Julien Lorion wrote: > Dear all, > > I am working on the evolution of deep-sea symbiotic mussels... I have got a > tree and 5 characters: habitat (hydrothermal vents, cold seep and organic > substrate), presence/absence of methanotrophic symbionts, presence/absence of > sulfoxydizing symbionts, symbiont location (extra VS intracellular) and body > length... > > So that's 1 continuous and 4 discrete binary variables (actually, I assumed > vent and seeps are very similar... so no need to take into account the 3 > states) > > At first I tested various hypotheses about correlation between my discrete > characters... I chose the easy way: I remembered my master lectures and used > basic Pagel's correlations. If you think that any new tool performs better, > I'd be happy to hear it. > > For now, my main concern is that I wanna test the impact of two binary > variables (habitat and symbiont location) on the body length... In fact, that > looks likes ANOVAs from which I wanna remove the phylogenetic bias... > > The point is that I don't know how to do that in practice. May someone have > some advices ? > > Thanks by advance > > Best regards > &
Re: [R-sig-phylo] nonparametric PGLS
I general, you should not be worrying about normality of the "tip" data, but rather the residuals of whatever multiple regression-type model you are implementing. Check those, and maybe they will be OK with a particular combination of independent variables. Bad residuals can adversely affect the likelihood of a model, so particularly check for big outliers. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Pascal Title [pascalti...@gmail.com] Sent: Thursday, May 10, 2012 8:46 PM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] nonparametric PGLS Hi everyone I have character data for species on a phylogeny, but simple transformations like log-transformations or square-root transformations are not proving to be sufficient to get the data to be normal. If this were a non-phylogenetic test, I would resort to something like a Spearman correlation, which is non-parametric. But with phylogenetic generalized least squares, is there any method that is nonparametric? Thanks! -Pascal Title [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger)
Read over the Blomberg et al. (2003) paper. K is intended for continuous-valued traits and/or those evolving similar to Brownian motion. You could report it if you wished, but I would add that caveat if you do. The randomization test should be robust in any case. Cheers, Ted From: Nina Hobbhahn [n.hobbh...@gmail.com] Sent: Wednesday, April 25, 2012 9:19 AM To: Theodore Garland Jr Cc: Alejandro Gonzalez; Hunt, Gene; Enrico Rezende; r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger) Thanks all for your helpful contributions! I will use phylosignal. Ted, I'm not sure I understand your last comment, "when the data are not though of as continuous-valued and/or evolving similar to Brownian motion". What do you mean by that? Also, are you suggesting that I report the presence/absence of phylogenetic signal, but not the value of the K statistic? Many thanks again, Nina On 2012-04-25, at 5:54 PM, Theodore Garland Jr wrote: > However, calculating a K statistic is strange when the data are not thought > of as continuous-valued and/or evolving similar to Brownian motion. The > randomization test is OK, however. > > Cheers, > Ted > > From: Alejandro Gonzalez [alejandro.gonza...@ebd.csic.es] > > Sent: Wednesday, April 25, 2012 8:46 AM > > To: Theodore Garland Jr > > Cc: Nina Hobbhahn; r-sig-phylo@r-project.org > > Subject: Re: [R-sig-phylo] Normality requirement for assessment of lambda > with phylosig (phytools) and fitContinuous (geiger) > > > > > > > > Hello, > > > > > > The library picante in R implements Blomberg et al (2003) K estimate, Liam's > phytools package does as well. Phytools has the added advantage, if I > remember correctly, of allowing users to estimate K including within species > variation. > > > > > > Cheers > > > > > > Alejandro > > > > > > > > > > > On 25, Apr 2012, at 5:29 PM, Theodore Garland Jr wrote: > > > > I would suggest the randomization test in Blomberg et al. (2003). This will > give a valid significance test of the null hypothesis of no phylogenetic > signal. By itself, it does not give a measure of the strength (or amount) of > phylogenetic signal. Not > sure if it is implented in r. If not, I can send our Matlab code. > > > > Cheers, > > Ted > > > > Theodore Garland, Jr. > > Professor > > Department of Biology > > University of California, Riverside > > Riverside, CA 92521 > > Office Phone: (951) 827-3524 > > Home Phone: (951) 328-0820 > > Facsimile: (951) 827-4286 = Dept. office (not confidential) > > Email: tgarl...@ucr.edu > > http://www.biology.ucr.edu/people/faculty/Garland.html > > > > Experimental Evolution: Concepts, Methods, and Applications of Selection > Experiments. 2009. > > Edited by Theodore Garland, Jr. and Michael R. Rose > > http://www.ucpress.edu/book.php?isbn=9780520261808 > > (PDFs of chapters are available from me or from the individual authors) > > > > > > From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] > on behalf of Nina Hobbhahn [n.hobbh...@gmail.com] > > Sent: Wednesday, April 25, 2012 1:55 AM > > To: r-sig-phylo@r-project.org > > Subject: [R-sig-phylo] Normality requirement for assessment of lambda with > phylosig (phytools) and fitContinuous (geiger) > > > > Dear fellow list users, > > > > I would like to assess the magnitude of phylogenetic signal in two sets of > continuous data. Set 1 contains numerous zeros and is therefore non-normal. > Set 2 contains very little variation and is non-normal due to > underdispersion. Given that both data sets are > largely immune to transformations to normality, I am wondering whether the > lambda estimates for untransformed data derived from phylosig and > fitContinuous will be meaningful? If not, can you recommend transformations > or other methods of phylogenetic-signal > assessment that would be preferable? > > > > Thank you very much, > > > > Nina > > > > > > > > Dr. Nina Hobbhahn > > Post-doctoral fellow > > Lab of Prof. S. D. Johnson > > School of Life Sciences > > University of KwaZulu-Natal > > Private Bag X01 > > Scottsville, Pietermaritzburg, 3201 > > South Africa ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger)
However, calculating a K statistic is strange when the data are not thought of as continuous-valued and/or evolving similar to Brownian motion. The randomization test is OK, however. Cheers, Ted From: Alejandro Gonzalez [alejandro.gonza...@ebd.csic.es] Sent: Wednesday, April 25, 2012 8:46 AM To: Theodore Garland Jr Cc: Nina Hobbhahn; r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger) Hello, The library picante in R implements Blomberg et al (2003) K estimate, Liam's phytools package does as well. Phytools has the added advantage, if I remember correctly, of allowing users to estimate K including within species variation. Cheers Alejandro On 25, Apr 2012, at 5:29 PM, Theodore Garland Jr wrote: I would suggest the randomization test in Blomberg et al. (2003). This will give a valid significance test of the null hypothesis of no phylogenetic signal. By itself, it does not give a measure of the strength (or amount) of phylogenetic signal. Not sure if it is implented in r. If not, I can send our Matlab code. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Nina Hobbhahn [n.hobbh...@gmail.com] Sent: Wednesday, April 25, 2012 1:55 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger) Dear fellow list users, I would like to assess the magnitude of phylogenetic signal in two sets of continuous data. Set 1 contains numerous zeros and is therefore non-normal. Set 2 contains very little variation and is non-normal due to underdispersion. Given that both data sets are largely immune to transformations to normality, I am wondering whether the lambda estimates for untransformed data derived from phylosig and fitContinuous will be meaningful? If not, can you recommend transformations or other methods of phylogenetic-signal assessment that would be preferable? Thank you very much, Nina Dr. Nina Hobbhahn Post-doctoral fellow Lab of Prof. S. D. Johnson School of Life Sciences University of KwaZulu-Natal Private Bag X01 Scottsville, Pietermaritzburg, 3201 South Africa [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo __ Alejandro Gonzalez Voyer Post-doc Estación Biológica de Doñana Consejo Superior de Investigaciones Científicas (CSIC) Av Américo Vespucio s/n 41092 Sevilla Spain Tel: + 34 - 954 466700, ext 1749 E-mail: alejandro.gonza...@ebd.csic.es Web site (Under construction): Personal page: http://consevol.org/members/alejandro.html Group page: http://consevol.org/index.html For PDF copies of papers see: http://csic.academia.edu/AlejandroGonzalezVoyer ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger)
I would suggest the randomization test in Blomberg et al. (2003). This will give a valid significance test of the null hypothesis of no phylogenetic signal. By itself, it does not give a measure of the strength (or amount) of phylogenetic signal. Not sure if it is implented in r. If not, I can send our Matlab code. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Nina Hobbhahn [n.hobbh...@gmail.com] Sent: Wednesday, April 25, 2012 1:55 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Normality requirement for assessment of lambda with phylosig (phytools) and fitContinuous (geiger) Dear fellow list users, I would like to assess the magnitude of phylogenetic signal in two sets of continuous data. Set 1 contains numerous zeros and is therefore non-normal. Set 2 contains very little variation and is non-normal due to underdispersion. Given that both data sets are largely immune to transformations to normality, I am wondering whether the lambda estimates for untransformed data derived from phylosig and fitContinuous will be meaningful? If not, can you recommend transformations or other methods of phylogenetic-signal assessment that would be preferable? Thank you very much, Nina Dr. Nina Hobbhahn Post-doctoral fellow Lab of Prof. S. D. Johnson School of Life Sciences University of KwaZulu-Natal Private Bag X01 Scottsville, Pietermaritzburg, 3201 South Africa [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Testing for phylogenetic signal in proportions
It is really too few species to have a sufficiently powerful test for phylogenetic signal. See Blomberg et al. (2003), available on my website. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 2009. Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Rob Lanfear [rob.lanf...@gmail.com] Sent: Wednesday, March 14, 2012 8:18 PM To: r-sig-phylo Subject: [R-sig-phylo] Testing for phylogenetic signal in proportions Hi All, I've tried searching the literature for this but the search terms tend to give irrelevant papers, so I thought I'd ask for some guidance here. I have a set of 11 species, an ultrametric tree for the species, and for each species I have a set of proportional data for 5 chemical traits. E.g. for spp1 my data might be: chemicalA=10%, chemicalB=50%, chemicalC=1%, chemicalD=19%, chemicalE=20%. The set of traits for each species always sums to 100%, but I don't know, and can't measure, the absolute values of the traits. Biologically speaking, this is OK, because it's the proportions that I'm interested in. I want to test for phylogenetic signal in these traits, and estimate the rate of change of each proportion along the phylogeny. Can anyone point me to any appropriate references for the methods and pitfalls of attempting to do this with proportion data? I can see that proportional data will have some odd properties (non-independence of traits, bounded (i.e. non-Brownian) evolution, etc.), but the best way of accounting for these is not immediately apparent to me. Thanks, Rob -- Rob Lanfear Research Fellow, Ecology, Evolution, and Genetics, Research School of Biology, Australian National University Tel: +61 2 6125 4321 www.robertlanfear.com [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Blomberg K Statistic
Hi Daniel, I see no reason NOT to calculate K and also to test for significance of the MSE. Your study design is an excellent one for testing the habitat effect (as we have pointed out in some papers). If habitat does have a strong effect, then you would expect the K value to be quite low. Note that K does not need to be near 1.00 in order for the randomization test for phylogenetic signal using the MSE to be statistically significant. Depending on sample size and so forth, many K statistics even below 0.5 can be associated with significant results in the randomization test for MSE (see Blomberg et al., 2003). All this said, I'd like to see your tree and some tip data -- can you send it as a Nexus file I can read into Mesquite? Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of dwe...@life.illinois.edu [dwe...@life.illinois.edu] Sent: Friday, January 20, 2012 9:38 AM To: liam.rev...@umb.edu Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] Blomberg K Statistic Hi Liam, Thanks for getting back to me. You are correct with #2- I have many bifurcations close to the present. The reason for this is that I am comparing between two different habitats of 21 different species. So, essentially, for each species, there are 2 tips (with branch lengths of 0)- 1 for each habitat (lentic and lotic). So, from what you are saying, I can't use the K statistic. Do you (or anyone else on this list) have an idea on a better way to test for phylogenetic significance on a tree where comparisons are made within species (between different habitats) as well as between species? Thank you for your thoughts. -Daniel > Hi Daniel. > > I can think of two possibilities: > > 1) This is an error. > > 2) Your tree has an extremely unusual shape (for instance, it has many > bifurcations very close to the present day). > > On constant rate Yule trees, substituting 1s for the true branch lengths > does not (on average) bias K in any particular direction. > > If you share your tree and data I would be happy to look at the issue > more closely. > > All the best, Liam > > -- > Liam J. Revell > University of Massachusetts Boston > web: http://faculty.umb.edu/liam.revell/ > email: liam.rev...@umb.edu > blog: http://phytools.blogspot.com > > On 1/19/2012 9:53 PM, dwe...@life.illinois.edu wrote: >> Hi, >>I'm using the Picante package in R to calculate the Blomberg K. I'm >> trying to test for a phylogenetic signal in my data. I'm using a >> dataset with the PC score for 21 different species on a tree with >> varying branch lengths. I did this using the code here: >> http://bodegaphylo.wikispot.org/IV._Testing_Phylogenetic_Signal_in_R >> >>I got a K statistic of 0.0006457603 and yet the resultant >> PIC.variance.P >> is 0.001. If I'm correct, the PIC.variance.P is the p-value and >> significant at 0.05. Is that correct? >>I then decided to play with my data and set all branch lengths equal >> to >> 1 to see how that would affect things. I got a K statistic of 2.921763 >> and a PIC.variance.P value of 0.001. My understanding is that K >> statistics close to 0 are not significant and become significant close >> to and above 1. >>I'm confused why I got a significant p-value for both K statistics, >> even >> though one is very close to 0 (0.00065) and the other is above 2. I >> would appreciate any help in explaining Picante and the K statistic. >> Thank you. >> >> >> >> >> -Daniel >> >> ><º> -><º> ><º> ><º> ><º> >> >> >> Daniel P Welsh >> PhD Candidate >> Teaching Assistant >> Department of Animal Biology >> University of Illinois at Urbana-Champaign >> 202 Shelford Vivarium >> 606 E. Healey Street >> Champaign, IL 61821 >> lab phone: (217) 333-5323 >> >> ___ >> R-sig-phylo mailing list >> R-sig-phylo@r-project.org >> ht
Re: [R-sig-phylo] R-sig-phylo Digest, Vol 48, Issue 12
I could not have said it better myself! Thanks, Luke! Ted From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Luke Matthews [lmatth...@activatenetworks.net] Sent: Thursday, January 19, 2012 6:41 AM To: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] R-sig-phylo Digest, Vol 48, Issue 12 Ana, I think what Ted and Paolo are getting at is you can't say the evolutionary relationships contributed to X% of the slope because the tree only deals with the nonindependence of the data points and is not itself an independent predictor variable. Because of this, regular gls regression applied to data that are nonindependent due to phylogeny produce unbiased estimates of the slope. The estimates of the standard error of the slopes, however, are biased to be too small when you run gls with nonindependent data. You are correct, however, that using a phylogeny will almost always alter the slope. This is because phylogenetic comparative methods (PCM) model the structure of the nonindependence and so change the slope estimate. This makes PCM and related methods (spatial models, network models) quite different from some methods of rank or robust regression that can produce exactly the same slope estimates as gls and will only increase the standard error estimates. In your particular case, the slope was higher. So it would be accurate to say, in this particular case, that ignoring the phylogeny resulted in a slope that was underestimated (not biased) by 26%. The phylogeny does not 'account for' this % of slope, but ignoring the phylogeny does produce a slope estimate that is inaccurate by 26% of the better PIC slope in this particular case. Best regards, Luke Luke J. Matthews Director of Data Analysis Activate Networks, Inc. Message: 3 Date: Wed, 18 Jan 2012 17:34:14 +0100 (CET) From: ppi...@uniroma3.it To: "Theodore Garland Jr" Cc: "r-sig-phylo@r-project.org" ,Ana Longo Subject: Re: [R-sig-phylo] Comparing slopes from regressions and PIC regressions Message-ID: <2397.79.32.224.88.1326904454.squir...@mail.uniroma3.it> Content-Type: text/plain;charset=iso-8859-1 VERY synthetically The two r-squared are NOT comparable. Taking in to account phylogeny is useful just to partial out data non independence due to shared ancestry, and to perform better tests of significance but the r-sq could be higher or smaller just because of chance. Maybe someone can add more on this. best paolo Long story short, you don't want to think about it in those terms. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Ana Longo [ana.lo...@gmail.com] Sent: Wednesday, January 18, 2012 7:20 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Comparing slopes from regressions and PIC regressions Dear all I am currently analyzing a community dataset of two continuous traits ? survival and disease resistance. I ran two regressions: (1) considering each species independent data points, and (2) phylogenetic independent contrasts. Both regressions were significant. Blomberg?s K also significant and ~ 0.38, and 0.34, respectively. However, I am interested in the slope of these regressions. For regression 1, slope = 0.142, for regression 2 slope = 0.105. I am interested in saying how much (in %) the evolutionary relationships contributed to the slope. My question is: Can I say that evolutionary relationships accounted for 74% (0.105/0.142) of the observed slope? Or is it only 26%? Thanks in advance for your help, since I don't have any experience running these types of analyses. Best regards, Ana [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo -- Paolo Piras Center for Evolutionary Ecology and Dipartimento di Scienze Geologiche, Universit? Roma Tre Largo San Leonardo Murialdo, 1, 00146 Roma Tel: +390657338000 email: ppi...@uniroma3.it ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Comparing slopes from regressions and PIC regressions
Long story short, you don't want to think about it in those terms. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Ana Longo [ana.lo...@gmail.com] Sent: Wednesday, January 18, 2012 7:20 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] Comparing slopes from regressions and PIC regressions Dear all I am currently analyzing a community dataset of two continuous traits – survival and disease resistance. I ran two regressions: (1) considering each species independent data points, and (2) phylogenetic independent contrasts. Both regressions were significant. Blomberg’s K also significant and ~ 0.38, and 0.34, respectively. However, I am interested in the slope of these regressions. For regression 1, slope = 0.142, for regression 2 slope = 0.105. I am interested in saying how much (in %) the evolutionary relationships contributed to the slope. My question is: Can I say that evolutionary relationships accounted for 74% (0.105/0.142) of the observed slope? Or is it only 26%? Thanks in advance for your help, since I don't have any experience running these types of analyses. Best regards, Ana [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] split data set and topology on PGLS
Why not just include an interaction term between your main effect coding variable (presumably coded as 0, 1 to indicate your two ecomorphs) and one or more of your continuous independent variables? Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Matt Pennell [mwpenn...@gmail.com] Sent: Wednesday, December 14, 2011 12:17 PM To: Renata Brandt Cc: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] split data set and topology on PGLS Hi Renata, This seems to be an appropriate approach. (I am sure someone will correct me if i am wrong in this). Phylogenetic regression in the way that you are using it is basically to remove the statistical non-independence between data points. Effectively what you are doing by splitting your data into two morphologies is pruning your tree. Under Brownian motion, evolution along any branch is independent of evolution along any other branch, so pruning your tree will not change the correlation struction of the data. (Similarily you can still do PGLS with incomplete sampling of species). So as long as you can justify splitting the data up as you are, it seems fine. cheers, matt On Wed, Dec 14, 2011 at 12:09 PM, Renata Brandt wrote: > Hello everybody. > > I have some questions for you. > When performing a phylogenetic regression between two continuous traits, > the dependent trait clearly divides my data set in two distinct > morphologies. This division overlaps with what I previously thought were > two different ecomorphs. The relationship between the same traits seems to > be different for each morphology. > > Now my question is: > - Is it ok to split the data set (one data table and topology for each > ecomorph) and reanalyze data to get the relationship between traits for > each group? (That is what I would do if not using a phylogenetic approach). > - If this is wrong, any hints on how should I proceed? > > Many thanks. Cheers. > > -- > Renata Brandt > Departamento de Biologia - FFCLRP > Universidade de São Paulo > Brasil > (16) 82039533 > > /..) > ( ( > ) ) > ( ( > ) ) > _( (_ > ) ) > \ ) >\ > >[[alternative HTML version deleted]] > > > ___ > R-sig-phylo mailing list > R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > > [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] Comparative Methods and Pseudo-Traits
Hi All, This is an interesting discussion. I'll draw your attention to two papers, one new and the other old. This is the new one: Grandcolas, P., R. Nattier, F. Legendre, and R. Pellens. 2011. Mapping extrinsic traits such as extinction risks or modelled bioclimatic niches on phylogenies: does it make sense at all? Cladistics 27:181-185. It is written from a somewhat different perspective than we usually have on this list. This is the old one, and I am taking the liberty of pasting in the relevant passage: Garland, T., Jr., P. H. Harvey, and A. R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41:18-32.Pages 29-30:WHAT KINDS OF TRAITSCAN BE ANALYZED? The independent contrasts approach isdesigned to investigate the correlated evolutionof traits that are inherited from ancestors,whatever the cause of that heritability.Thus, the phenotypic data for tipspecies are generally assumed to reflect underlyinggenetic differences among species,as could be verified through common-garden experiments (Garland and Adolph,1991). The tip species for one set of contrastscan even be different from those forthe other set, as long as the phylogenetictrees are isomorphic. This realization makesit possible to use independent contrasts toexamine coevolution of phenotypic traitssuch as body size and life span in coevolvedhost-parasite systems (see Harveyand Keymer, 1991). In addition to the usual phenotypic traits(e.g., body size, metabolic rate), cultural(see Cavalli-Sforza and Feldman, 1981), environmental(e.g., soil or water pH, meanannual temperature), and other traits thatare difficult to categorize (e.g., home rangearea) can be studied as long as they arepassed on from ancestral to descendentspecies (or populations) and have a continuousdistribution. For example, many environmentalproperties, such as latitude ormean annual rainfall, are not inherited inthe conventional (genetic) sense. Nevertheless,they are inherited in the sense thatorganisms are born into environmentalconditions and locations experienced bytheir parents at the time of birth. Thus, theancestor of two species living in a desertmay also have lived in a desert (cf. Huey,1987), or the ancestor of one high-latitudeand one equatorial species may have livedat midlatitude. Similarly, if an environmentalcharacteristic is determined solely(without externally imposed constraints)through a process of habitat selection, andif species differences in habitat selectionare genetically based, then species differencesin the environmental trait will begenetically based as well. Alternatively, ifvariation in some (genetically based) phenotypictrait can be used as a precise indicatorof some environmental characteristic,then that phenotypic trait may be usedas a surrogate for the environmental characteristic.For example, toe fringes in lizardsmight indicate occupancy of sandyhabitats. Unfortunately, this is not unfailinglythe case; some species that glidethrough the air or that run across wateralso possess toe fringes (Luke, 1986). Finally,paleoclimatological and historicalbiogeographical data might be used in conjunctionto indicate environmental characteristicsof hypothetical ancestral (as opposedto tip) species, but this takes us intothe realm of other comparative methods,such as those based on minimum evolutionreconstructions of ancestors (Huey, 1987;Harvey and Pagel, 1991; Maddison, 1991;Martins and Garland, 1991). In any case,techniques for correlating phenotypes withenvironmental characteristics require furtherstudy. Cheers,Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of David Bapst [dwba...@uchicago.edu] Sent: Monday, November 14, 2011 12:54 PM To: Liam J. Revell; Joe Felsenstein; pasquale.r...@libero.it Cc: R Sig Phylo Listserv Subject: Re: [R-sig-phylo] Comparative Methods and Pseudo-Traits Liam, Joe, Pasquale, all- Thank you for your kind input.It seems that I am not the only one who considers this issue at length. There is just one point I'd like clarification of. Liam, in my first example which you used, the inherited trait is the response and the not-directly-inheritable trait the predictor,
Re: [R-sig-phylo] Is correlation of PICs, with tip data and each terminal node split into male and female, a valid method?
Your power to detect a correlation is also going to be low. For example, see the figure in Garland and Adolph (1994), or any standard power curve for a correlation coefficient. If you are comparing two models (e.g., OLS and PGLS) with the same number of independent variables in them (e.g., one), the you don't need AIC. Just look at the likelihoods. Bigger is better. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Alberto Gallano [alberto@gmail.com] Sent: Friday, November 04, 2011 2:10 PM To: R-phylo Mailing-list Subject: Re: [R-sig-phylo] Is correlation of PICs, with tip data and each terminal node split into male and female, a valid method? Dear Marguerite and Ted, thanks you very much, your suggestions make sense to me - I am looking at the references you mentioned. I'd like to ask for one small clarification regarding power. I know from several papers that power is not good below n=17 for independent contrasts, as you mentioned. What I want to know is, does my small sample size affect the actual significance of the correlation (i.e., does it reduce my ability to get a significant p-value for the correlation)? Or, as I think is more likely, does it just reduce my ability to tell whether the phylogenetic model is a better fit than an ordinary statistical model? By the way, how would I assess the model fit? If I was using pGLS the functions would generate an AIC value for each model. But using the cor.table or other correlations functions, no AIC value is produced. Is there a way to judge model fit when doing PIC correlations? thank you se much, Alberto On Thu, Nov 3, 2011 at 1:22 PM, Theodore Garland Jr < theodore.garl...@ucr.edu> wrote: > Dear Alberto, > > I agree with all of what Marguerite wrote. Here are two good papers to > check on the metrics: > > Abouheif, E., and D. J. Fairbairn. 1997. A comparative analysis of > allometry for sexual size dimorphism: assessing Rensch's rule. Am. Nat. > 149:540-562. > > Cox, R. M., S. L. Skelly, and H. B. John-Alder. 2003. A comparative test > of adaptive hypotheses for sexual dimorphism in lizards. Evolution > 57:1653-1669. > > And, with so few species, you are not going to be able to tell very well > whether a phylogenetic statistical model, using a hierarchical tree, fits > your data better than a conventional analysis, assuming a star phylgony, so > present both. > > Cheers, > Ted > > Theodore Garland, Jr. > Professor > Department of Biology > University of California, Riverside > Riverside, CA 92521 > Office Phone: (951) 827-3524 > Home Phone: (951) 328-0820 > Facsimile: (951) 827-4286 = Dept. office (not confidential) > Email: tgarl...@ucr.edu > http://www.biology.ucr.edu/people/faculty/Garland.html > > Experimental Evolution: Concepts, Methods, and Applications of Selection > Experiments > Edited by Theodore Garland, Jr. and Michael R. Rose > http://www.ucpress.edu/book.php?isbn=9780520261808 > (PDFs of chapters are available from me or from the individual authors) > > > From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] > on behalf of Marguerite Butler [mbutler...@gmail.com] > Sent: Thursday, November 03, 2011 9:49 AM > To: Alberto Gallano > Cc: R-phylo Mailing-list > Subject: Re: [R-sig-phylo] Is correlation of PICs, with tip data and > each terminal node split into male and female,a valid method? > > Dear Alberto, > > I think the problem with including both males and females as separate > datapoints in the analysis is that you're artificially doubling the sample > size, which will make your statistical results difficult to interpret. > > Why don't you do two separate analyses (or three)? One with males only, > one with females only, and one on sexual dimorphism (some sort of > difference or ratio or log-ratio variable -- people have used various > metrics in the literature)? > > Doing comparative analyses on 10 species is going to have very low power, > no matter what you do, however. > > Good luck! > Marguerite > > On Nov 2, 2011, at 4:29 PM, Alberto Gallano w
Re: [R-sig-phylo] Is correlation of PICs, with tip data and each terminal node split into male and female, a valid method?
Dear Alberto,
I agree with all of what Marguerite wrote. Here are two good papers to check
on the metrics:
Abouheif, E., and D. J. Fairbairn. 1997. A comparative analysis of allometry
for sexual size dimorphism: assessing Rensch's rule. Am. Nat. 149:540-562.
Cox, R. M., S. L. Skelly, and H. B. John-Alder. 2003. A comparative test of
adaptive hypotheses for sexual dimorphism in lizards. Evolution 57:1653-1669.
And, with so few species, you are not going to be able to tell very well
whether a phylogenetic statistical model, using a hierarchical tree, fits your
data better than a conventional analysis, assuming a star phylgony, so present
both.
Cheers,
Ted
Theodore Garland, Jr.
Professor
Department of Biology
University of California, Riverside
Riverside, CA 92521
Office Phone: (951) 827-3524
Home Phone: (951) 328-0820
Facsimile: (951) 827-4286 = Dept. office (not confidential)
Email: tgarl...@ucr.edu
http://www.biology.ucr.edu/people/faculty/Garland.html
Experimental Evolution: Concepts, Methods, and Applications of Selection
Experiments
Edited by Theodore Garland, Jr. and Michael R. Rose
http://www.ucpress.edu/book.php?isbn=9780520261808
(PDFs of chapters are available from me or from the individual authors)
From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on
behalf of Marguerite Butler [mbutler...@gmail.com]
Sent: Thursday, November 03, 2011 9:49 AM
To: Alberto Gallano
Cc: R-phylo Mailing-list
Subject: Re: [R-sig-phylo] Is correlation of PICs, with tip data and each
terminal node split into male and female,a valid method?
Dear Alberto,
I think the problem with including both males and females as separate
datapoints in the analysis is that you're artificially doubling the sample
size, which will make your statistical results difficult to interpret.
Why don't you do two separate analyses (or three)? One with males only, one
with females only, and one on sexual dimorphism (some sort of difference or
ratio or log-ratio variable -- people have used various metrics in the
literature)?
Doing comparative analyses on 10 species is going to have very low power, no
matter what you do, however.
Good luck!
Marguerite
On Nov 2, 2011, at 4:29 PM, Alberto Gallano wrote:
> Dear useRs,
>
> I am sorry if my previous question was not clear - my English speaking
> colleague helped me reword my question:
>
> I want to correlate a body size proxy to another (shape) variable using
> independent contrasts. *But*, but I want to split the data by sex (as there
> is much sexual dimorphism in body size in the species I study). This means
> I need to have a tree where each terminal species is actually made up of
> two nodes (male and female). I assume to do this I must make the branch
> lengths between the sexes within each species tiny (e.g., 1e-7), so that
> they are basically contemporaneous (it doesn't seem to work if branch
> lengths between sexes are set to zero). Is this the best way to apply
> correlation with PICs to each sex? Is there a problem with having each
> terminal taxon split by sex? (see tree object below)
>
> My variables are "bodySize" (a geometric mean proxy) and "var1" a 'shape'
> index - log10(var1 / bodySize). I want to look for an allometric effect in
> "var1" by correlating with "bodySize", after correcting for isometric size
> differences. My real data is for 10 species, but here I give an example
> with 3 species and each sex. I am mainly using Spearman's rho, as the small
> sample has many outliers, but I give example functions for Pearson's r as
> well below.
>
> thank you,
>
> Alberto
>
>
> # -
> # dummy datas
> set.seed(186)
> dat <- data.frame(bodySize=rnorm(6, 10), var1=rnorm(6, 5))
> rownames(dat) <- c("taxon1_M", "taxon1_F", "taxon2_M", "taxon2_F",
> "taxon3_M", "taxon3_F")
>
> # tree
> library(ape)
> tree <- structure(list(edge = structure(c(7L, 8L, 9L, 9L, 8L, 10L, 10L,
> 7L, 11L, 11L, 8L, 9L, 1L, 2L, 10L, 3L, 4L, 11L, 5L, 6L), .Dim = c(10L,
> 2L)), Nnode = 5L, tip.label = c("taxon1_M", "taxon1_F", "taxon2_M",
> "taxon2_F", "taxon3_M", "taxon3_F"), edge.length = c(3.4, 2.399,
> 1e-07, 1e-07, 2.399, 1e-07, 1e-07, 5.799, 1e-07, 1e-07
> ), root.edge = 0, tip.labels = c("taxon1_M", "taxon1_F", "taxon2_M",
> "taxon2_F", "taxon3_M", "taxon3_F")), .Names = c("edge", "Nnode",
> "tip.label", "edge.length", "root.edge", "tip.labels"), class = "phylo")
>
>
> # -
Re: [R-sig-phylo] specifying weights in gls
I think you might want to be using one over the square root of the sample size as the weights. We have sometimes done this when using phylogenetically independent contrasts: Bonine, K. E., and T. Garland, Jr. 1999. Sprint performance of phrynosomatid lizards, measured on a high-speed treadmill, correlates with hindlimb length. Journal of Zoology, London 248:255-265. Perry, G., and T. Garland, Jr. 2002. Lizard home ranges revisited: effects of sex, body size, diet, habitat, and phylogeny. Ecology 83:1870-1885. This paper also has some relevant things: Ives, A. R., P. E. Midford, and T. Garland, Jr. 2007. Within-species variation and measurement error in phylogenetic comparative methods. Systematic Biology 56:252-270. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Will Stein [rwst...@sfu.ca] Sent: Thursday, September 22, 2011 1:07 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] specifying weights in gls Hi All, I am working on a comparative analysis involving a relatively small taxa set (16), one continuous dependent character, one continuous independent character, and a dummy variable (0,1) that specifies group membership (two groups). I've been running bivariate regressions with and without the dummy variable in APE using gls and corPagel, and an ultrametric phylogeny that is well resolved. To the issue: the species-specifc mean values that compose the continuous independent variable were generated from a wide array of sample sizes (range 4 - 300). My impression is that there is heterogeneity in the precision of the estimates resulting from sample size variability and that both groups are similarly heterogeneous. I would like to account for this heterogeneity in the independent variable explicitly; however, I do not have the variances of the species-specific mean values, only the sample size they were generated from, which suggests using a weight that is proportional to the sample size. I've looked into the varFunc and the associated varClass functions available in gls. My initial impression is that varFixed seems an appropriate varClass option and I am getting reasonable results with varFixed(~1/sample.size). I am wondering if others have dealt with this issue directly, and if so, how? This must be a relatively common consideration. The model set up ends being: weighted.gls.fit <- gls(Dep.Var ~ Indep.Var, weights=varFixed(~1/sample.size),correlation=corPagel(0.5,phy), data=data.frame) summary(weighted.gls.fit) Best, Will Stein ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] squared-change parsimony reconstruction
I have not tried this, but why should you need to set the branch lengths equal to one? That is not a requirement of the squared-change parsimony algorithm (or ML equivalent) per se. (With variable branch lengths, people often refer to it as weighted squared-change parsimony.) Maybe this is just a limitation in ACE? Also, how does ACE handle polytomies? I know that some algorithms have trouble with polytomies, which may be indicated as zero-length branches. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Sam Brown [s_d_j_br...@hotmail.com] Sent: Friday, August 26, 2011 5:11 PM To: r-sig-phylo@r-project.org; onmik...@gmail.com Subject: Re: [R-sig-phylo] squared-change parsimony reconstruction >> Date: Thu, 25 Aug 2011 13:13:47 +0200 >> From: Ond?ej Mikula >> To: r-sig-phylo@r-project.org >> Subject: [R-sig-phylo] squared-change parsimony reconstruction >> Message-ID: >> >> Content-Type: text/plain >> >> Dear all, >> >> I would like to make squared-change parsimony reconstruction of ancestral >> states for continuous multivariate data as recommended by: Hi Ondrej This can be done by setting the branch lengths of your tree to '1' and running ace(vector_of_states, your_tree, type = "continuous", method = "ML") See also the R-phylo wiki: http://www.r-phylo.org/wiki/HowTo/Ancestral_State_Reconstruction Cheers! Samuel Brown [[alternative HTML version deleted]] ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] R: Re: R: Re: R: ancestral state reconstruction for tips
Hi Pas, No worries, we have all done an accidental "Reply All" more than once! >I estimated ancestral (and tip) values for which I have real data via >BM assumption to see how good the fit is Can you clarify? Unless you have some a priori hypothesis to test about a particular tip (or set of tips, such as a whole clade), then why would you estimate their values and how would you do this? Did you just delete one at a time, crank the numbers (presumably yielding the same values as you would get from Garland and Ives, 2000), and see what you got? Whether it is a "tip" or a fossil taxon (which is just a tip with a branch that terminates before now), the confidence intervals on the predicted values will be hugely affected by how long the branch is to that taxon (the longer the branch, the wider the prediction intervals). And also by how many "close relatives" are attached to the node it comes from, and by how much phenotypic diversity exists in those "close relatives." >The bottom line is answering the question: how long should the branch >leading to that particular species be if it evolved at the same rate of its >sister species? That's an interesting way to look at it (a sort of inverse [perverse?] parameterization), but it does not give you any additional information beyond asking whether a taxon is an "outlier" via the tests we have discussed a bit ago. Or am I missing something? Cheers, Ted From: pasquale.r...@libero.it [pasquale.r...@libero.it] Sent: Friday, August 05, 2011 12:38 PM To: j...@gs.washington.edu Cc: dwba...@uchicago.edu; hu...@si.edu; Theodore Garland Jr; r-sig-phylo@r-project.org Subject: R: Re: R: Re: [R-sig-phylo] R: Re: R: ancestral state reconstruction for tips Folks, I was intending my most recent message to be off-list and didn't realize "r-sig-phylo@r-project.org" was in the CC field, which means I'm a fool. All kidding aside, yes Joe, I estimated ancestral (and tip) values for which I have real data via BM assumption to see how good the fit is. Actually, estimated values are very close to real values for some species, barely so for some others, and absolutely not for others still. The good news is that since there is a single mode of evolution tree wise, deviations from real values really mean that evolution is accelerated, or decelerated, either, in these particular lineages for which a significant deviation from the expected value is noticeable. What I', trying to do now is writing a R routine to back-calculate the "expected" branch lengths for the "unusual" critters, given the fitted ancestral values and tip values of the phenotypes, and assuming BM, in order to compare the actual branch lengths to the expected. The ratio of these ! lengths, if I'm not delusional and definitely lucky, is a per-lineage rate of phenotypic evolution. The bottom line is answering the question: how long should the branch leading to that particular species be if it evolved at the same rate of its sister species? Pas Messaggio originale Da: j...@gs.washington.edu Data: 05/08/2011 21.04 A: "pasquale.r...@libero.it" Cc: , , , "r-sig-phylo@r-project.org" Ogg: Re: R: Re: [R-sig-phylo] R: Re: R: ancestral state reconstruction for tips Folks -- I was intending my most recent message to be apologetic -- that I was perhaps overreactive. Certainly Pas has not raised unreasonable objections or been obstructive with my grants! (Others have). Let me raise an issue so I understand him more clearly: Pas, are you saying that you see phenotypes in the fossils that seem incompatible with the Brownian Motion assumption? Joe Joe Felsenstein j...@gs.washington.edu Dept of Genome Sciences and Dept of Biology, Univ. of Washington, Box 5065, Seattle Wa 98195-5065 ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] R: Re: R: ancestral state reconstruction for tips
>Having said that, my guess was that we *may* use the BM and >computations at nodes to see where (in which lineages) do >phenotypes appear very different from predictions. >For instance, I think it could be somehow possible to use >estimated ancestral charactes to see how much the inclusion >of some fossil (or new) species changes the estimated value >(e.g. by creating a polytomy by the inclusion of the new species), >or even back-calculate branch lengths (under BM assumption) >for these unusual phenotypes to see how much evolution >accelerates in these lineages (by comparison with real branch lengths). Yes, I think that is an excellent way to proceed. Cheers, Ted From: pasquale.r...@libero.it [pasquale.r...@libero.it] Sent: Friday, August 05, 2011 12:00 PM To: dwba...@uchicago.edu; j...@gs.washington.edu; hu...@si.edu; Theodore Garland Jr Cc: r-sig-phylo@r-project.org Subject: R: Re: [R-sig-phylo] R: Re: R: ancestral state reconstruction for tips Hi All, I'm happy I have stimulated some discussion about this subject matter. For some reason I can't imagine it looks this whole thing is going to be somehow personal and I have not posted this last e-mail to the list as a consequence. Joe, unfotunately I never attended a lecture of yours, and didn't raise trivial distinctions and objections to a grant proposal you submitted. My intention was not to be critical about BM or ICs, or whatever. I just wanted to point it out that things are sometimes a bit too complex and some unreliable predictions from our models may slip out unnoticed evey now and then, as I believe it is apparent reading the literature (including my own, of course). Having said that, my guess was that we *may* use the BM and computations at nodes to see where (in which lineages) do phenotypes appear very different from predictions. For instance, I think it could be somehow possible to use estimated ancestral charactes to see how much the inclusion of some fossil (or new) species changes the estimated value (e.g. by creating a polytomy by the inclusion of the new species), or even back-calculate branch lenghts (under BM assumption) for these unusual phenotypes to see how much evolution accelerates in these lineages (by comparison with real branch lengths). I hope I spoke my mind more clearly at this time. Pas >Messaggio originale >Da: dwba...@uchicago.edu >Data: 05/08/2011 20.23 >A: "Joe Felsenstein" >Cc: "r-sig-phylo@r-project.org" >Ogg: Re: [R-sig-phylo] R: Re: R: ancestral state reconstruction for tips > >As the diversity of explicit models of trait evolution grow, it will >be interesting to see if any consensus develops about which models >hold most often in general and whether any insight is gained into >which conditions predict appearance of different models. > >I think Joe is right that realizing a model is an inaccurate or >imprecise description of reality should impel us to develop better >models of the world around us, because this partly how science moves >forward. However, I don't think pointing out that a model is deficient >requires that that person must themselves develop an alternative. >After all, an alternative model that capture a more realistic level of >complexity may not be possible in some situations (it is certainly >possible in trait evolution models, however.) Requiring such a thing >would put too much pressure on scientific whistle-blowers, who play a >very important role in reminding the rest of us that the world is more >than the models we use to understand it and make our predictions. > >-Dave > > > > >On Fri, Aug 5, 2011 at 10:51 AM, Joe Felsenstein wrote: >> >> Pasquale Raia said: >> >>> Of course Ted is right, but my problem with this computation, or >>> with the >>> simple exercise I was proposing is well another: as a >>> paleontologist I often >>> come across pretty exceptional phenotypes (dwarf hippos and >>> elephants, huge >>> flightless birds, to make a few examples). When you use methods >>> like this (I >>> mean Garland and Ives') and compare the output with those >>> phenotypes, as I did, >>> you immediately realize what the the bottom line is: no matter if >>> they are >>> nodes or tips, by using the expected (under BM) covariance the >>> estimated >>> phenotypes are dull, perfectly reasonable but very different from >>> anything >>> exceptional you may find yourself to work with. This is why I feel >>> it is >>> difficult to rely on those (unobserved) values to begin with. >> >> I think that what is being said is that Brownian Motion is too sedate >> a process >> and does not predict some of the large
Re: [R-sig-phylo] R: ancestral state reconstruction for tips
Hi Pas, I have not followed all of the background on this, so am not sure exactly what you want to do. The relevant methods that were developed in Garland and Ives (2000, e.g., Figs. 2 and 3), and also shown earlier in a slightly different formulation here (see Fig. 4): Garland, T., Jr., and S. C. Adolph. 1994. Why not to do two-species comparative studies: limitations on inferring adaptation. Physiological Zoology 67:797-828. were intended mainly for testing hypotheses about whether a single species (of a priori interest) deviates significantly from, say, an allometric relationship. For that purpose, I think the use of these methods and assuming Brownian-motion like character evolution makes sense. The nice thing about these procedures from a paleontological perspective is that you can, if appropriate, "hang" the species of interest anywhere on the tree, including coming off of internal nodes by (very) short branch lengths. If there is something different that you are trying to do, then I'd be happy to discuss it further. Maybe offline with pictures? Cheers, Ted From: pasquale.r...@libero.it [pasquale.r...@libero.it] Sent: Friday, August 05, 2011 9:41 AM To: Theodore Garland Jr; Hunt, Gene; r-sig-phylo@r-project.org Subject: R: Re: [R-sig-phylo] R: ancestral state reconstruction for tips Of course Ted is right, but my problem with this computation, or with the simple exercise I was proposing is well another: as a paleontologist I often come across pretty exceptional phenotypes (dwarf hippos and elephants, huge flightless birds, to make a few examples). When you use methods like this (I mean Garland and Ives') and compare the output with those phenotypes, as I did, you immediately realize what the the bottom line is: no matter if they are nodes or tips, by using the expected (under BM) covariance the estimated phenotypes are dull, perfectly reasonable but very different from anything exceptional you may find yourself to work with. This is why I feel it is difficult to rely on those (unobserved) values to begin with. Any opinion? Pas >Messaggio originale >Da: theodore.garl...@ucr.edu >Data: 05/08/2011 18.24 >A: "Hunt, Gene", "r-sig-phylo@r-project.org" >Ogg: Re: [R-sig-phylo] R: ancestral state reconstruction for tips > >The methods in the Garland and Ives (2000) paper are in our package of DOS PDAP programs, and should also be functional in the PDAP module of Mesquite. > >Cheers, >Ted > >Theodore Garland, Jr. >Professor >Department of Biology >University of California, Riverside >Riverside, CA 92521 >Office Phone: (951) 827-3524 >Home Phone: (951) 328-0820 >Facsimile: (951) 827-4286 = Dept. office (not confidential) >Email: tgarl...@ucr.edu >http://www.biology.ucr.edu/people/faculty/Garland.html > >Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments >Edited by Theodore Garland, Jr. and Michael R. Rose >http://www.ucpress.edu/book.php?isbn=9780520261808 >(PDFs of chapters are available from me or from the individual authors) > > >From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Hunt, Gene [hu...@si.edu] >Sent: Friday, August 05, 2011 8:35 AM >To: r-sig-phylo@r-project.org >Subject: Re: [R-sig-phylo] R: ancestral state reconstruction for tips > >Also, the issue of predicting values for unknown tips using data from other species in the tree is considered in this reference: > >Garland, T., and A. R. Ives. 2000. Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. American Naturalist 155(3):346-364. > >Best, >Gene > > > >On 8/5/11 11:31 AM, "pasquale.r...@libero.it" wrote: > > > > >Hi Morgan, > >this is just stuff for thought, and remember, this is wrong anyway. But you >may try something like this: > >1. compute pics, >2. take the pic value at the ancestral node subtending to your unknown tip, >3. pretend one of the two tips the pic was originally computed on is in fact >your unknown species, >4. modify the square of the summed branch lengths of the two species by using >the "new" bl, >5. use the formula for pics (standardized) to derive your unknown tip value by >using the other (real) species tip value and the new square of summed branch >lengthts > > >but again, remember this is wrong, because contrasts were computed without >your unknown species. With ace everything turns out to be much more complicated >because ancestral value estimations are 'optimized' by taking the entire tree >and distribution of values at once, so to speak. > > > > > >>Messaggio or
Re: [R-sig-phylo] R: ancestral state reconstruction for tips
The methods in the Garland and Ives (2000) paper are in our package of DOS PDAP programs, and should also be functional in the PDAP module of Mesquite. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Hunt, Gene [hu...@si.edu] Sent: Friday, August 05, 2011 8:35 AM To: r-sig-phylo@r-project.org Subject: Re: [R-sig-phylo] R: ancestral state reconstruction for tips Also, the issue of predicting values for unknown tips using data from other species in the tree is considered in this reference: Garland, T., and A. R. Ives. 2000. Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. American Naturalist 155(3):346-364. Best, Gene On 8/5/11 11:31 AM, "pasquale.r...@libero.it" wrote: Hi Morgan, this is just stuff for thought, and remember, this is wrong anyway. But you may try something like this: 1. compute pics, 2. take the pic value at the ancestral node subtending to your unknown tip, 3. pretend one of the two tips the pic was originally computed on is in fact your unknown species, 4. modify the square of the summed branch lengths of the two species by using the "new" bl, 5. use the formula for pics (standardized) to derive your unknown tip value by using the other (real) species tip value and the new square of summed branch lengthts but again, remember this is wrong, because contrasts were computed without your unknown species. With ace everything turns out to be much more complicated because ancestral value estimations are 'optimized' by taking the entire tree and distribution of values at once, so to speak. >Messaggio originale >Da: morgan.g.i.langi...@gmail.com >Data: 05/08/2011 14.15 >A: >Ogg: [R-sig-phylo] ancestral state reconstruction for tips > >I was wondering if there is a way to get ancestral state >reconstructions not for nodes within the tree but for tips that I >don't know the trait of. I could do this somewhat manually, by taking >the ancestral state resconstruction from the parent and child nodes >surrounding where my unknown tip branches off from the tree and >averaging those results (weighted by the branch length). This approach >seems kind of clunky, so I was hoping there was something better. > > > >Morgan Langille > >___ >R-sig-phylo mailing list >R-sig-phylo@r-project.org >https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo -- Gene Hunt Curator, Department of Paleobiology National Museum of Natural History Smithsonian Institution [NHB, MRC 121] P.O. Box 37012 Washington DC 20013-7012 Phone: 202-633-1331 Fax: 202-786-2832 http://paleobiology.si.edu/staff/individuals/hunt.cfm ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] pic() vs gls()
The independent contrasts algebra for what Joe is talking about can be found in these papers: Garland, T., Jr., P. E. Midford, and A. R. Ives. 1999. An introduction to phylogenetically based statistical methods, with a new method for confidence intervals on ancestral values. American Zoologist 39:374-388. Garland, T., Jr., and A. R. Ives. 2000. Using the past to predict the present: Confidence intervals for regression equations in phylogenetic comparative methods. American Naturalist 155:346-364. Cheers, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of Joe Felsenstein [j...@gs.washington.edu] Sent: Thursday, July 14, 2011 6:35 AM To: ppi...@uniroma3.it Cc: r-sig-phylo Subject: Re: [R-sig-phylo] pic() vs gls() Paolo Piras wrote -- > Citing > http://www.r-phylo.org/wiki/HowTo/Ancestral_State_Reconstruction: > Using Felsenstein's (1985) phylogenetic independent > contrasts (pic); This is also a Brownian-motion based > estimator, but it only takes descendants of each node > into account in reconstructing the state at that node. > More basal nodes are ignored. > > I THINK that, on the opposite, more basal nodes are > NOT ignored in gls and for this reason results can > differ slightly > I'm wrong? The contrast algorithm if continued to the root, makes the correct ancestral reconstruction for the root. You are correct that values for higher nodes in the tree are not the correct ML reconstruction for those nodes. If the tree is rerooted at any interior node and the algorithm used for that, then that node's reconstruction will be correct. There are ways of re-using information so that the total effort of doing this for all interior nodes will be no worse than about twice that of a single pass through the tree. However people may prefer to use PGLS, which if properly done should give the proper estimates for all nodes. There is some discussion of this in Rohlf's 2001 paper in Evolution. Joe Joe Felsenstein, j...@gs.washington.edu Dept. of Genome Sciences, Univ. of Washington Box 355065, Seattle, WA 98195-5065 USA ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] ancestral state reconstruction with fixed internal node(s)
Hi Graham,
I, too, have done this under Brownian motion and "squared-change parsimony"
reconstructions.
In my experience (which was a while ago, and memory fades ...), the confidence
intervals for nodes near to the one you constrain ("near" in terms of the
length of the branches connecting the nodes) should be reduced, and sometimes
substantially, if the constrained node is close by. Right?
Cheers,
Ted
Theodore Garland, Jr.
Professor
Department of Biology
University of California, Riverside
Riverside, CA 92521
Office Phone: (951) 827-3524
Wet Lab Phone: (951) 827-5724
Dry Lab Phone: (951) 827-4026
Home Phone: (951) 328-0820
Facsimile: (951) 827-4286 = Dept. office (not confidential)
Email: tgarl...@ucr.edu
http://www.biology.ucr.edu/people/faculty/Garland.html
Experimental Evolution: Concepts, Methods, and Applications of Selection
Experiments
Edited by Theodore Garland, Jr. and Michael R. Rose
http://www.ucpress.edu/book.php?isbn=9780520261808
(PDFs of chapters are available from me or from the individual authors)
From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on
behalf of Graham Slater [gsla...@ucla.edu]
Sent: Thursday, June 30, 2011 10:09 AM
To: r-sig-phylo@r-project.org
Subject: Re: [R-sig-phylo] ancestral state reconstruction with fixed internal
node(s)
Hi AnneMarie,
The way Paolo suggests would be the best/right way to do this. We're
working on some methods for incorporating fossil info in comparative
methods and I have some code that will do this that i can send you
off-list if you would like. I should add that, at least based on what
we've found so far, doing this probably won't make much difference to
your reconstructed ancestral states, simply because under BM, your CIs
are so wide without constraining nodes that doing so has little effect.
best,
Graham
On 06/30/2011 09:40 AM, ppi...@uniroma3.it wrote:
> Hi AnneMarie,
> I dont really know if this makes sense; in fact
> ancestral state reconstruction is an ** a posteriori
> estimation ** of nodal values starting from tips
> observations.
>
> A trick could be to add a false taxon lnked to that
> node and giving to it a 0 branch length (i.e.
> plitomized - you can then resolve this during
> computation using multi2di() ) and assigning your
> desired trait value.
>
> I'm not sure if this helps
>
> Best
> Paolo
>
>
>
>
>
>
> Dear phylo-sig list people,
>
> I want to do ancestral state reconstruction
> (preferably with ace) with
> one (or more) of the internal nodes 'fixed' for a
> range / a distribution
> of values. For instance, I want a node leading to one
> particular clade
> that is present a subset of my trees to have a value
> from 0.2-0.5, and
> then do the ancestral state reconstruction with this
> restriction on that
> node.
>
> I have been searching the archives and the net for
> discussion of this
> (but maybe with the wrong search terms), and I cannot
> find anything
> about it - not how to do it in R or in any other
> package.
>
> Thanks for your input!
> Annemarie
>
> --
> Annemarie Verkerk, MA
> Evolutionary Processes in Language and Culture (PhD
> student)
> Max Planck Institute for Psycholinguistics
> P.O. Box 310, 6500AH Nijmegen, The Netherlands
> +31 (0)24 3521 185
> http://www.mpi.nl/research/research-projects/evolutionary-processes
>
> ___
> R-sig-phylo mailing list
> R-sig-phylo@r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
>
> ___
> R-sig-phylo mailing list
> R-sig-phylo@r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
___
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
___
R-sig-phylo mailing list
R-sig-phylo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
Re: [R-sig-phylo] small sample size and detecting phylogenetic signal with Blombergs K statistic
Dear Charlotte, No, you do not want to do this. As shown in Blomberg et al. (2003), the tests for statistical significance do not work well when you have fewer than about 20 species. The estimate of the K statistic (indicating the amoung of phylogentic signal) is fine for small sample sizes, but the test of its statistical significance is not. You can compare your value of K against the many other values reported in our paper to see if it is "typical" for that type of trait. Note that K can range from close to zero to somewhat grater than one for real data. See the values in Blomberg et al. (2003). Typical values are in the range of about 0.4 to 0.8, and those values are tyupically significantly greater than zero when you have decent sample size (greater than 20 or so). A value a bit below one does NOT mean no phylogenetic signal. In general, it is better to use statistical models that simultaneously estiumate the strength of phylogenetic signal and the other parameters (e.g., partial regression slopes). However, unfortunately, this also does not work well with small samples sizes. So, I suggest you do both conventional statistical analyses and phylogenetically independent contrasts (which is equivalent to PGLS models without any sort of branch length transform). Hopefully, results will be similar. If not, then you probably cannot conclude much from your data. Sincerely, Ted Theodore Garland, Jr. Professor Department of Biology University of California, Riverside Riverside, CA 92521 Office Phone: (951) 827-3524 Wet Lab Phone: (951) 827-5724 Dry Lab Phone: (951) 827-4026 Home Phone: (951) 328-0820 Facsimile: (951) 827-4286 = Dept. office (not confidential) Email: tgarl...@ucr.edu http://www.biology.ucr.edu/people/faculty/Garland.html Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments Edited by Theodore Garland, Jr. and Michael R. Rose http://www.ucpress.edu/book.php?isbn=9780520261808 (PDFs of chapters are available from me or from the individual authors) From: r-sig-phylo-boun...@r-project.org [r-sig-phylo-boun...@r-project.org] on behalf of charlotte De Busschere [charlotte.debussch...@ugent.be] Sent: Thursday, June 30, 2011 8:45 AM To: r-sig-phylo@r-project.org Subject: [R-sig-phylo] small sample size and detecting phylogenetic signal with Blombergs K statistic Dear all, I'm currently describing phenotypic divergence patterns of several traits within a radiated genus (5 species). Several comparative studies suggest to take into account the phylogenetic relationships among the species by means of PIC or PLGS etc. for which I agree But before applying these methods I guess I should first detect whether there is Phylogenetic signal (PS)in the data and if not then I suppose I can use classic statistical methods (correct me if my reasoning is wrong) Hence I calculated Blombergs K statistics with Picante and almost all traits have K values<1 i.e. no PS (but no p values < 0.05) hence my question concerns about the accuracy of this statistic in a small data set (N=5) Blomberg et. al. 2003 (Evolution) mentioned that it is practically impossible to get a statistical significant result if N<7 Hence, can I put confidence in this statistic? Sincerely greetings, Charlotte Drs. Charlotte De Busschere Ghent University Biology Department Terrestrial Ecology Unit Ledeganckstraat 35 B-9000 Gent Belgium tel:++32 (0)92645039 http://www.ecology.ugent.be/terec/personal.php?pers=cd ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo ___ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
