My data set consist of number of calls (lcin) across Day. I am looking for activity differences between three features (4 sites per feature). I am also looking for peaks of activity across time (Day). I am using a gamm since I believe these are nonlinear trends with nested data. gammdata<-gamm(lcin~Temp+s(Day)+fType+wind+fFeature+Forest+Water+Built, list=fSite,data=data, family=gaussian) summary(gammdata$gam) summary(gammdata$lme) anova(gammdata$gam) I can see which variables are significant but I was wondering if there was a way to do a post hoc to see differences between features? Or is there a way to compare different models (feature) for significance (not best fit) against each other? Thanks, Rham -- View this message in context: http://r.789695.n4.nabble.com/Gamm-and-post-comparison-tp4411366p4411366.html Sent from the R help mailing list archive at Nabble.com.
Gamm and post comparison
2 messages · RHam, Gavin Simpson
6 days later
On Wed, 2012-02-22 at 10:55 -0800, RHam wrote:
My data set consist of number of calls (lcin) across Day. I am looking for activity differences between three features (4 sites per feature). I am also looking for peaks of activity across time (Day). I am using a gamm since I believe these are nonlinear trends with nested data.
You don't need gamm() for this as you don't appear to be using any random effects. gam() will be fine for the model you show below. I don't know enough about the workings of mgcv to know whether you can leverage functions in the multcomp package; gam objects inherit from glm and lm classes and multcomp can work with these, though do not the point in ?gamObject that gam objects lack the details of the fitting expected in glm or lm objects. IIRC, that model could be fitted as a linear mixed model if the smooth term is set up correctly. multcomp can work with nlme and lme4 so you could continue to use gamm() as below but use the multcomp package on the $lme component. The gamm4 package provides a version of gamm() that uses lme4 instead of nlme for the underlying fitting. Again you might look into using the lmer representation of the model with multcomp. None of the above is tested or even based on personal experience of doing this. G
gammdata<-gamm(lcin~Temp+s(Day)+fType+wind+fFeature+Forest+Water+Built, list=fSite,data=data, family=gaussian) summary(gammdata$gam) summary(gammdata$lme) anova(gammdata$gam) I can see which variables are significant but I was wondering if there was a way to do a post hoc to see differences between features? Or is there a way to compare different models (feature) for significance (not best fit) against each other? Thanks, Rham -- View this message in context: http://r.789695.n4.nabble.com/Gamm-and-post-comparison-tp4411366p4411366.html Sent from the R help mailing list archive at Nabble.com.
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~% Dr. Gavin Simpson [t] +44 (0)20 7679 0522 ECRC, UCL Geography, [f] +44 (0)20 7679 0565 Pearson Building, [e] gavin.simpsonATNOSPAMucl.ac.uk Gower Street, London [w] http://www.ucl.ac.uk/~ucfagls/ UK. WC1E 6BT. [w] http://www.freshwaters.org.uk %~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%