2

I have a generalized linear mixed model with a three way interaction, a nested random variable, and a binomial response variable :

modelB15=glmer(cbind(resistant, (total-resistant) )~
               time*inoc_source*inoc_resistance+
               (1|block/plot),
               family = binomial,
               data = B15)

the corresponding matrix to generate linear combinations of coefficients (a contrast matrix) is:

time1_sourceHY_mix      = c(1,0,0,0,0,0,0,0,0,0,0,0)
time2_sourceHY_mix      = c(1,1,0,0,0,0,0,0,0,0,0,0)
time1_sourceNV_mix      = c(1,0,1,0,0,0,0,0,0,0,0,0)
time1_sourceHY_negative = c(1,0,0,1,0,0,0,0,0,0,0,0)
time1_sourceHY_positive = c(1,0,0,0,1,0,0,0,0,0,0,0)
time1_sourceNV_negative = c(1,0,1,1,0,1,0,0,0,0,0,0)
time1_sourceNV_positive = c(1,0,1,0,1,0,1,0,0,0,0,0)
time2_sourceNV_mix      = c(1,1,1,0,0,0,0,1,0,0,0,0)
time2_sourceHY_negative = c(1,1,0,1,0,0,0,0,1,0,0,0)
time2_sourceHY_positive = c(1,1,0,0,1,0,0,0,0,1,0,0)
time2_sourceNV_negative = c(1,1,1,1,0,1,0,1,1,0,1,0)
time2_sourceNV_positive = c(1,1,1,0,1,0,1,1,0,1,0,1)

I then can pull up the vectors to test different hypotheses, such as "is time 1 different from time 2"? 

time_HYpositive = time2_sourceHY_negative-time1_sourceHY_positive
time_HYnegative = time2_sourceHY_negative-time1_sourceHY_negative
time_HYmix      = time2_sourceHY_mix-time1_sourceHY_mix
time_NVpositive = time2_sourceNV_negative-time1_sourceNV_positive
time_NVnegative = time2_sourceNV_negative-time1_sourceNV_negative
time_NVmix      = time2_sourceNV_mix-time1_sourceNV_mix

timeDiff=rbind(time_HYpositive, time_HYnegative, time_HYmix,
 time_NVpositive,time_NVnegative,time_NVmix)

I want to use estimable() in library(gmodels) to make different comparisons between means for each coefficient or combination of coefficients: 

estimable(modelB15, timeDiff, conf.int=.991)

but I get the following error message: 

Error in FUN(newX[, i], ...) : 'param' has no names and does not match number of coefficients of model. Unable to construct coefficient vector

estimable() works great with a glm, but a) how can it be modified to work with lmer() b) if estimable won't work with a GLMM, what other functions could accomplish the same task?

I have tried glmmPQL(), it doesn't work either.

The author in this page is trying to accomplish a similar thing.

This page was a bit helpful.

Community
  • 1
  • 1
C.H.Hagerty
  • 21
  • 4

1 Answers1

0

estimable() will not work with glmer()

However, the lsmeans() package is compatible with generalized linear mixed models and can generate LS means via pairwise comparisons. This gives the user the ability to compare combinations of coefficients, without generating a messy matrix.

[Rethinking the analysis of non-normal data in plant and soil science] by Walter W. Stroup has an excellent description and supplementary documents contain useful example R code

C.H.Hagerty
  • 21
  • 4