[R-sig-ME] MCMCglmm for unbalanced panel data

[Wenjian Jia]

Dear list,

I am modeling number of annual registered battery electric vehicle (BEV)
and plug-in hybrid electric vehicle (PHEV) in 133 counties in US. I have
unbalanced panel data that some counties have four years of observations,
while some counties only have one or two years observations. I plan to use
the MCMCglmm package to estimate a bivariate log-normal Poisson mixed model:

   1. County is the random effect
   2. BEV and PHEV are two dependent variables
   3. X1 - X5 are independent variables.

model <- MCMCglmm(cbind(BEV, PHEV) ~ trait-1  + trait:X1 + trait:X2 +
> trait:X3 + trait:X4 + trait:X5,
>              random = ~ us(trait):COUNTY,  rcov = ~ us(trait):units,
>              family = rep("poisson", 2), prior = prior, nitt = 200000,
> thin = 100, burnin = 15000, verbose = FALSE,  pr = TRUE, data = data)
>

My questions is: are the code above also appropriate for my unbalanced
dataset? Thanks in advance!

Best,
-- 
Wenjian Jia, PhD student
Department of Civil and Environmental Engineering
University of Virginia

	[[alternative HTML version deleted]]