Cecile and Joe, Thank you for the replies. That helped clear up a lot for me.
And thank you for the link Cecile. Estimating sigma_phy is quite a pain because of all the matrix inversions. I've a tree of approx. 2000 tips, so even using of a Cholesky decomposed V_{phy} doesn't buy me much time at all. Because the vector of means for the MVN in my work is all 0s, I've written my own MVN sampler to try and minimize how many inversions or determinants need to be calculated. The stan mailing list had a great little discussion on how to speed up estimation of MVN random effects with partially known covariance matrices https://groups.google.com/forum/#!topic/stan-users/HcvYaDu71_Y and I just followed that code. Now my posterior sampling at least gets off the ground and finishes. The stan list discussion possibly provides a useful starting template for other people on this list trying to do similar things. Cheers, Peter On Mon, Dec 22, 2014 at 11:32 AM, Cecile Ane <a...@stat.wisc.edu> wrote: > Hi Peter, > > We have taken a similar approach in this paper: > http://onlinelibrary.wiley.com/doi/10.1111/evo.12582/abstract > We also use a Bayesian approach, with a prior that allows us to integrate > the ancestral state and the total variance analytically (no need for MCMC > for those parameters). Our notation for your sigma_phy is lambda, like > Pagel's, and we do use multiple individuals per species. For what it's > worth. > > Regarding the scaling of branch lengths, I agree with Joe that there is > nothing particular about 1, other than providing an easier interpretation > for the numerical value of the phylogenetic variance. > Cheers, > Cecile. > > > On 12/14/2014 01:03 PM, Peter Smits wrote: > >> Hi list, >> >> I have a similar question to Edwin. I too am working with a hierarchical >> Bayesian model, though I've implemented it in stan. I've included a >> phylogenetic random effect term, which is modeled as being distributed as >> multivariate normal with known covariance matrix up to a constant, >> sigma_{phy}. This follows Lynch '91 Am Nat and Housworth et al. '04 Am Nat >> by drawing on the similarity with the "animal model" from quantitative >> genetics. >> >> My question is about the scaling of the covariance matrix: is it necessary >> to have the the diagonal terms satisfy x <= 1 and all the off diagonals to >> be 0 <= x < 1? I have a time scaled phylogeny from which I have my >> covariance matrix, so currently all elements of the matrix are not scaled >> so that the greatest distance from the root to a tip is 1. Currently, the >> elements of the matrix are just the sum of the shared branch lengths. Is >> this appropriate? Why or why not? >> >> Any input would be much appreciated. >> >> Cheers, >> >> Peter Smits >> > > > -- > Cecile Ane > Departments of Statistics and of Botany > University of Wisconsin - Madison > www.stat.wisc.edu/~ane/ > > CALS statistical consulting lab: > www.cals.wisc.edu/calslab/stat_consulting.php > > > _______________________________________________ > 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-ph...@r-project.org/ > -- Peter D Smits Grad student Committee on Evolutionary Biology University of Chicago psm...@uchicago.edu http://home.uchicago.edu/~psmits/home.html [[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/