Hi R users, I have some uncertanties regarding comparative method in R. Regarding especially the lamba and K values. I am using 'ape' and 'nlme' packages and 'corPagel' and 'gls' functions in a regression model. My questions regards this model. If I wish to calculate signal in multiple regression (say two variables) I can use formula: Y ~ X + bodymass. This will take account of body mass. What If I only want to find signal in one variable? Should the model be: Y ~ bodymass ? Will this give lambda value for Y after controling for body mass, or should I instead correct for body mass using a ratio (Y / body mass) and get lambda for this - using perhaps 'fitContinuous' in 'geiger'. It seems better to me to use regression model, but I welcome your thoughts on the matter.
I also whish to calculate K value. This can be done with 'phylosignal' function in 'picante' package. This calculates K for one variable (already scaled for body mass). Is it possible to calculate K in regression model, like lambda? One final question - I see people use often logarithm of trait values in pgls. Is this ok? It seems to me this transformation would reduce distance between means along the right tail of a sample and would thus distort signal measurment? kind regards, Alberto [[alternative HTML version deleted]] _______________________________________________ R-sig-phylo mailing list R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo