[R-sig-ME] lmer and post-hoc testing

SHONA MARY LESLIE 1001632L at student.gla.ac.uk
Mon Jan 10 17:54:22 CET 2011

Hi there,

I am investigating habitat preference in 9-spined sticklebacks using 4 experimental arenas (each fish spent same time in each arena) and want to know whether the time in the shallow area was different between arenas.

I carried out a generalised linear mixed model (using lmer), putting in the fish ID as a random effect, and the distribution as binomial (since it's proportion data); as follows...

> model1<-lmer(y~factor(Arena)+SL+(1|FishID), family=binomial)
> summary(model1)

This worked but I can't work out how to run a Tukey test, to see which arenas were different.  I have tried... (with error message underneath)

> TukeyHSD(model1)
Error in UseMethod("TukeyHSD") : 
  no applicable method for 'TukeyHSD' applied to an object of class "mer"

> summary(glht(model1,linfct=mcp(Arena="Tukey"))) 
Error in summary(glht(model1, linfct = mcp(Arena = "Tukey"))) : 
  error in evaluating the argument 'object' in selecting a method for function 'summary'

> glht(model1, linfct=mcp(Arena="Tukey"))
Error in mcp2matrix(model, linfct = linfct) : 
  Variable(s) ‘Arena’ have been specified in ‘linfct’ but cannot be found in ‘model’! 

> glht(model1, linfct=mcp(factor(Arena)="Tukey")
Error: unexpected '=' in "glht(model1, linfct=mcp(factor(Arena)="

I have read some of the R help archives and it has been pointed out that Tukey may or may not work with lmer (I downloaded the newest version and still no luck).  Any ideas?  If Tukey won't work, are there others that might?


More information about the R-sig-mixed-models mailing list