Does anyone know how to undertake Generalized Estimating Equation (GEE) modelling using the negative binomial distribution in R?

I have monthly count data on health outcome X and I have count data on exposure Y over 10 years. I have one covariate age in categories. Both X and Y have seasonal clustering. Tried to use poisson regression and data seemed overdispersed. Used negative binomial and GEE with poisson distribution.

I set an experiment where different individuals of the same spider species are offered different prey. All the spiders receive all the prey once and we count the immobilization times against them. In order to analyze this data, I was thinking about applying a GLMM, nevertheless I have received some suggestions about using GEE (General Estimation Equations). I have tried to look for some sources about when to apply what analysis, but there is no information. I would like to know if any of you know when it is better to apply each analysis.

We know the generalized linear models (GLMs) are a broad class of models. When fitting GLMs in R, we need to specify which family function to use from a bunch of options like gaussian, poisson, binomial, quasi, etc. I would like to have your advice regarding how to determine the optional family function used for GLM fitting in R. Thanks!

I have two models having the same fixed effects - one assumes the correlations of the repeated measures is autoregressive corAR1 while the assumes the correlation structure is independent. How can I compare the two models?

I built a GLMM using glmer() from the package "lme4" to conduct a poisson regression. However, overdispersion was detected and the family "poisson" therefore cannot be used. Moreover, "quasipoisson" families are not supported by the glmer function. Can anyone suggest a solution for this problem? Many thanks for your attention!

I have a pre-post treatment and comparison quasi-experiment for business foundings and failures.I have been able to use xtnbreg or xtgee in Stata to estimate this regression, but I would like to be able to do it in R. It would also be nice to have zero-inflated models. When I consulted experts last year, the options included cross sectional GEE in R using Poisson or quasi-poisson. Or to use glm.nb cross-sectional analysis. There is a pglm package, but it does not have the GEE population averaged model. As far as I can tell, neither gee nor geepack take the negative binomial family from the MASS package. However, I see a new package geeM and it looks like it might work. Does anyone have experience using it?

My experimental design includes;

• One fixed factor (2 levels)

• An experimental unit is a tank containing 8 subjects

• For each treatment condition a random factor, n = 5 (i.e. 5 tanks each containing 8 subjects)

• The response variable is the proportion of subjects (n/8) that have abandoned a shell

• For each tank/experimental unit, measures were taken 10 times (i.e., non-independent measures because; P @ t+2 is dependent on t+1).

I have tried the following script – but I’m having difficulty getting at the random “tank” factor

>Model<-glm(cumulativeP ~ time + treat + time*treat + (1|tank), family=quasibinomial, data = data1)

Any suggestions and/or examples of how to interpret the output would be greatly appreciated.