Hi Viviane, If you believe that your traits are correlated - that is there is a variance-covariance matrix that describes the evolutionary correlations among these traits, then I dont think you don't want to fit the models to each trait and transform the tree each time you simulate. Youd be better off estimating lambda for all traits simultaneously. You can do this in Gavin Thomas' motmot package http://gavinhthomas.wordpress.com/motmot/, and Im sure Liam has something to do this in phytools too. You can then use the trait VCV and your multivariate lambda to simulate correlated traits on an appropriately scaled tree. You could do this directly (drawing from a multivariate normal distribution after taking the Kronecker product of your phylogenetic and trait VCVs) or else you can use a function like geigers sim.char a trait vcv matrix to generate correlated traits.
More importantly, I dont think you want to simulate with a lambda transform to generate your null model if youre interested in knowing whether your traits are over or underdispersed. If your multivariate trait lambda is indeed close to 1, then the phylogenetic covariances among taxa do a relatively good job of predicting similarity and your trait is not really overdispersed relative to a constant rate process like Brownian motion. I think what you really want to do is generate a VCV for the traits and then simulate correlated trait evolution under BM on your unscaled tree. You can then use some measure of dispersion in multivariate space to ask whether your observed trait variance is significantly lower or higher than expected under BM. Best, 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 (202) 633-1316 slat...@si.edu<mailto:slat...@si.edu> www.fourdimensionalbiology.com<http://www.fourdimensionalbiology.com> On Aug 7, 2014, at 7:40 AM, Viviane Deslandes <viviane.deslan...@gmail.com<mailto:viviane.deslan...@gmail.com>> wrote: I am investigating patterns of phenotypic overdispersion/clustering in a reginal scale. I am using a set of traits to calculate phenotypic diversity. I need to simulate this set of correlated traits to generate a null model. I would like to know whether the following approach is appropriate: - I used fitContinuous::Geiger to verify which model of evolution best fits to my traits.The models used were White Noise, Ornstein Uhlenbeck, Brownian Motion, Early Bust and lambda. - I compared the models according to its AICc values. Most of my traits fitted better to lambda model with values next to 1 (0.9). Can I transform the original tree based on lambda values for each trait separately (using transform.phylo) and then to use a function to simulate the evolution of each song trait? Which is the appropriate function to simulate ( "RtraitCont", "sim.corrs" or fastBM)? Thank you, Viviane [[alternative HTML version deleted]] _______________________________________________ R-sig-phylo mailing list - R-sig-phylo@r-project.org<mailto: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/