Following Dormann et al 2007 Ecography, I have employed a GLMM approach in R to account for spatial autocorrelation in a binomial regression model (logistic regression) that does not have random terms. Using data from mtcars (just so we all have the same numbers), my code looks as follows:
library(MASS) data <- mtcars data$group <- factor(rep("A", nrow(data))) mod1 <- glmmPQL(vs ~ mpg, random=~1|group, correlation=corExp(form=~disp+qsec), data=data, family=binomial)
Question 1: Is this a reasonable way to account for spatial autocorrelation in a binomial model, as Dormann et al suggest?
Question 2: Should/can I demonstrate that there is no longer spatial autocorrelation (SAC) in the error for this model? Can a vario/correlagram be used? How? I have had trouble finding information about how to look for SAC in GLMMs, and the variogram I've managed to create looks very funky -- hard to interpret.