Hi Sebastien, I believe the simulation simply draws random numbers from the appropriate multivariate normal distribution determined by the model's covariance matrix for the tip values. This means that a complete trajectory of trait values in never actually generated. This is much more efficient than simulating the stochastic differential equation step by step along each branch, and achieves the same result.
To obtain simulated trait values at the internal nodes, you would again draw from the multivariate normal distribution, but you would want a covariance matrix which was n_nodes by n_nodes, not just n_tips by n_tips. This covariance matrix is relatively straightforward to compute. For instance, in the case of Brownian motion you would first calculate the Brownian rate parameter sigma. The covariance matrix is just sigma^2 times the matrix of divergence times. The matrix of divergence times has entry i,j equal to the length of time from the root until those two nodes diverged (age of their common ancestor measured from the root). (So if they shared no evolutionary history, they have no covariance, and if i=j, the the time is equal to the time since the root, sigma^2 *t, the expected variance). Using this matrix you would just draw from the multivariate normal distribution with this as the covariance matrix. Don't know if code for this exists already (anyone?) but we could write something up to do this if you like. Sounds useful. Hope that made sense, Carl On Wed, Aug 4, 2010 at 6:59 AM, Sebastien Lavergne < sebastien.laver...@ujf-grenoble.fr> wrote: > I forgot to mention that I am willing to do this with continuous traits. > Thanks to those who already answered (about discrete characters > unfortunately). > Seb > > Le 04/08/2010 15:03, Sebastien Lavergne a écrit : > > Hi everybody, >> >> I have a question regarding the sim.char function in geiger. >> Does anybody foresee an easy way to retrieve the simulated ancestral >> states ? >> (i.e. the values of each trait for all internal nodes of the tree) >> Does this sound doable ? Any hint about how to do it ? >> >> Thanks for your help >> Seb >> >> >> > -- > ------------------------------------------------------------------------- > Sébastien Lavergne > Laboratoire d'Ecologie Alpine, UMR-CNRS 5553 > Université Joseph Fourier > BP 53, 38041 Grenoble Cedex 9, France > tel +33 (0)4 76 63 54 50 > http://www-leca.ujf-grenoble.fr/membres/lavergne.htm > > _______________________________________________ > R-sig-phylo mailing list > R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > -- Carl Boettiger Population Biology, UC Davis http://two.ucdavis.edu/~cboettig [[alternative HTML version deleted]]
_______________________________________________ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo
