I`m fitting a "poisson" MCMC genetic model using MCMCglmm. My response variable is clearly poisson distributed, constituted by count data (integers) ranging from 1 to 120.
I`ve already ran simmilar models in ASREML and variance components range from 0.01 to 0.4, including animal, permanent environment effects and residual.
The question is:
I`ve read in many MCMCglmm documents that priors for G structure are often set to G1=list(V=1, nu=0.002) and the same for the R structure.
But thinking about my specific problem, initial values for the variance of 1 wouldn`t be correct, right? If yes, what would be a good prior for small variance components?
Am I missinterpreting V and nu ? (which in correct notation would be alpha and beta for the inverse-gamma distribution)
- Badges
- Science topic