[R] vis.gam() plot graph of differences??

Mabille, Geraldine Geraldine.Mabille at nina.no
Wed Jun 6 13:25:31 CEST 2012


Dear all,
I'm using gam models to test for a by-factor interaction and then plotting the results using vis.gam(). The y variable is binomial and I use models of this sort:
Mod1<-gam(y~Factor+te(x1,x2, by=Factor) +s(Group,bs="re"),data=mydata,family=binomial(link=logit),method="REML")
Mod2<-gam(y~Factor+te(x1,x2)+te(x1,x2,by=as.numeric(Factor==1))+ te(x1,x2,by=as.numeric(Factor==2))+s(Group,bs="re"),data=mydata,family=binomial(link=logit),method="REML")
With Mod1, I use vis.gam to visualize the 2D-smooth obtained for each level of my Factor (3 levels):
vis.gam(Mod1, view=c("x1","x2"),cond=list(Factor=0),plot.type="contour",color="heat")
vis.gam(Mod1, view=c("x1","x2"),cond=list(Factor=1),plot.type="contour",color="heat")
vis.gam(Mod1, view=c("x1","x2"),cond=list(Factor=2),plot.type="contour",color="heat")
For Mod2 however, I try to plot with vis.gam() the 2D-smooth for the reference level and then the 2D-smooths showing the difference between reference level (0) and level 1, and the difference between reference level(0) and level 2. Those are the three plots usually obtained when using plot(Mod2) but that I would like to plot with vis.gam() in order to add nice colors to the plot.
I have tried several options but don't manage to get those plots with vis.gam(). Does anyone has any idea of some coding I could write to obtain those??
Thanks if anyone can help,
Geraldine



More information about the R-help mailing list