----- Forwarded message from Chris Klingenberg -----
Date: Thu, 4 Oct 2012 16:53:12 -0400
From: Chris Klingenberg
Reply-To: Chris Klingenberg
Subject: Re: MorphoJ update
To: [email protected]
Dear Andrea
The bug fixes concerning the insufficient degrees of freedom require no
re-runs: in the old version, they caused crashes. In the new version,
MorphoJ will refuse to do the analyses (we are talking situations when
there are no degrees of freedom).
The bug fixes concerning independent contrasts in the presence of
zero-length branches require a rerun if you have a phylogeny that
contains branches of length zero. (That's the cautious version; in
reality, the bug only struck if the phylogeny contained one or more
internal nodes where both descendant branches were of length zero.) This
should be a relatively rare situation.
The situation about discriminant analysis and cross-validation is
related to the story with the degrees of freedom. If the samples are
small relative to the number of landmarks, the total sample size is less
than the number of dimensions, and the within-group degrees of freedom
are therefore less than the dimensionality of the shape space. In this
situation, discriminant analysis makes no sense (any two sets of
observations can be "perfectly" discriminated -- try it with a triangle
in 2D....). Well-behaved software should refuse to do the analysis with
a polite error message.
I should include this in the next series of bug fixes for MorphoJ.
Thanks. But with any software, including MorphoJ, users should think
whether their data are sufficient for doing an analysis... This includes
users with semilandmarks in their data.
I'm not sure how the 100% wrongly classified observations in the
cross-validation are coming about.But we are talking more or less ugly
nonsense, but definitely nonsense.
I hope this helps.
Best wishes,
Chris
On 04/10/2012 15:44, [email protected] wrote:
> ----- Forwarded message from [email protected] -----
>
> Date: Wed, 03 Oct 2012 14:17:02 -0700
> From: [email protected]
> Reply-To: [email protected]
> Subject: Re: MorphoJ update
> To: [email protected]
>
> ----- Forwarded message from [email protected] -----
>
> Date: Sat, 29 Sep 2012 23:16:55 -0700
> From: [email protected]
> Reply-To: [email protected]
> Subject: Re: MorphoJ update
> To: [email protected]
>
> ----- Forwarded message from [email protected] -----
>
> Date: Fri, 28 Sep 2012 18:20:25 -0700
> From: [email protected]
> Reply-To: [email protected]
> Subject: Re: MorphoJ update
> To: [email protected]
>
> ----- Forwarded message from andrea cardini -----
>
> Date: Fri, 28 Sep 2012 04:24:40 -0400
> From: andrea cardini
> Reply-To: andrea cardini
> Subject: Re: MorphoJ update
> To: [email protected]
>
> Dear Chris,
> when a project is reloaded, will the latest version redo the computations
> of the old analyses? Or will they have to be redone manually one by one?
>
> On a potentially related issue to one of those you fixed, I've had
> colleagues reporting some apparently strange results (100% wrongly
> classified individuals in x-validated DAs) using small samples relative to
> the number of landmarks. That did not happen, I believe, when we re-did
> those analyses in R, PAST, SPSS using PCs (all of them, subsets and also, I
> think, we might have done it using simulated data). I wonder whether that
> too was some kind of issue with small sample sizes.
> I guess you'd say that one should not do DAs in those cases, and I agree,
> but yet this might happen more often now that many people use
> semilandmarks. Possibly the software could issue a warning.
>
> Thanks for your help and for going on improving MorphoJ. It's a wonderful
> software.
> Cheers
>
> Andrea
>
> At 13:50 26/09/2012 -0700, you wrote:
>> ----- Forwarded message from Chris Klingenberg -----
>>
>> Date: Wed, 26 Sep 2012 14:32:13 -0400
>> From: Chris Klingenberg
>> Reply-To: [email protected]
>> Subject: MorphoJ update
>> To: [email protected]
>>
>> Dear morphometricians
>>
>> Yesterday, I uploaded a new release of MorphoJ, version 1.05c, to the
>> web site:
>> http://www.flywings.org.uk/MorphoJ_page.htm
>>
>> This version includes bug fixes in several places in the program that
>> stop analyses if there are not enough degrees of freedom, and prevents
>> crashes in those analyses of if the output is later used otherwise. It
>> affects the following: Generate Covariance Matrix, Regression, Partial
>> Least Squares (both between separate blocks and within a single
>> configuarion of landmarks). Thanks to Soledad de Esteban and the
>> participants of the recent workshop on morphometrics and phylogeny in
>> Sabadell for making me aware of the problem.
>>
>> The program also fixes a bug in the computation of independent contrasts
>> (when mapping data onto a phylogeny), which concerned the handling of
>> zero-length branches. Thanks to Soledad de Esteban for telling me about
>> the problem.
>>
>> A further minor bug fix was added in the computation of PC scores from
>> phylogenetically independent contrasts. For this kind of data, PC scores
>> are now computed without centering (a contrast of zero will now produce
>> a PC score of zero).
>>
>> Best wishes,
>> Chris
>>
>> --
>> ***************************************************************
>> Christian Peter Klingenberg
>> Faculty of Life Sciences
>> The University of Manchester
>> Michael Smith Building
>> Oxford Road
>> Manchester M13 9PT
>> United Kingdom
>>
>> Telephone: +44 161 275 3899
>> Fax: +44 161 275 5082
>> E-mail: [email protected]
>> Web: http://www.flywings.org.uk
>> Skype: chris_klingenberg
>> ***************************************************************
>>
>> ----- End forwarded message -----
>>
>>
> Dr. Andrea Cardini
> NEW ADDRESS: Dipartimento di Scienze Chimiche e Geologiche, Università di
> Modena e Reggio Emilia, l.go S. Eufemia 19, 41121 Modena, Italy
> FORMERLY: Dipartimento di Biologia, Universitá di Modena e Reggio Emilia,
> via Campi 213, 41100, Modena, Italy
> tel: 0039 059 2055017 ; fax: 0039 059 2055548
>
> Centre for Anatomical and Human Sciences
> University of Hull, Cottingham Road, Hull, HU6 7RX, UK
> University of York, Heslington, York YO10 5DD, UK
>
> Centre for Forensic Science , The University of Western Australia
> 35 Stirling Highway, Crawley WA 6009, Australia
>
> E-mail address: [email protected], [email protected]
> Webpage: http://sites.google.com/site/hymsfme/drandreacardini
>
> Editorial board for:
> Zoomorphology:
> http://www.springer.com/life+sciences/animal+sciences/journal/435
> Journal of Zoological Systematics and Evolutionary Research:
> http://www.wiley.com/bw/journal.asp?ref=0947-5745&site=1
> Hystrix, the Italian Journal of Mammalogy:
> http://www.italian-journal-of-mammalogy.it/
>
> ----- End forwarded message -----
>
> ----- End forwarded message -----
>
> ----- End forwarded message -----
>
> ----- End forwarded message -----
>
>
--
***********************************
Christian Peter Klingenberg
Faculty of Life Sciences
University of Manchester
Michael Smith Building
Oxford Road
Manchester M13 9PT
United Kingdom
E-mail: [email protected]
Phone: +44 161 2753899
Skype: chris_klingenberg
***********************************
----- End forwarded message -----
