[R-sig-ME] G-computation from GLMM fitted values

[benton at stat.berkeley.edu]

Hi:

I am trying to estimate the uncertainty of a G-computation estimated from
the fitted values of a GLMM, specifically, a binomial model with one
random effect and several fixed effects.  I'd like to obtain the variance
of the mean difference in proportions between treatment groups by
bootstrapping the GLMM and taking the variance of these difference across
many bootstrap samples.  It seems that the LME4a bootMer function would
work, but I do not understand the "FUN" part of this function.  Your help
is greatly appreciated.

Best,

Katie