Hi Sereina, The two models are not equivalent. The first one is fitting only an interaction effect while the second one fits the main effects plus the interaction term. Something like this:
Model 1 y~x1:x2 Model 2 y~x1+x2+x1:x2 As far as I know in general only model 2 makes sense and I would not try to fit model 1, but there may be circumstances in which model 1 would make sense. Hard to say without knowing the question and data. Best Kaspar On Friday, 18 July 2014, Sereina Graber <sereina.gra...@gmx.ch> wrote: > Dear all, > > I have question concerning the interaction effects in PGLS. Using several > different data sets and questions, I have tried to inlcude interaction > effects in pgls regression models for several times now, however, I have > never had a significant interaction effect, even though it is pretty clear > that is should be significant. But, if I include the product of two > variables as a new variable into the pgls model, this shows a strong effect > (code see below). > Does anyone has an idea what might be the problem here? And would it be ok > to simply include the product of the two predictor variables as a new > variable into an analysis instead of actually implementing the interaction > ? > > Product of two predictors as a new variable: > X_prod<-X1 * X2 > model1<-pgls(Y ~ X_prod, comp_data) > > ...instead of : > model1<-pgls(Y ~ X1 * X2, comp_data) > > Thank you & best, > Sereina > -- Kaspar Delhey ARC DECRA research fellow School of Biological Sciences, Monash University 3800, Clayton, Victoria, Australia phone: +61-(0)3-9902 0377 skype: kaspar.delhey web: https://sites.google.com/site/kaspardelhey/ [[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/
