Dear list,

I am experimenting with proc glimmix in SAS v. 9.3, and I wonder how to interpret the generalized chi-square statistic that is reported as a measure of model fit? I have tried Gamma and Inverse Gaussian distributions for multilevel modeling of very skewed variables, and SASreports a chi-square and a chi-square/df statistic. Is it just "smaller is better", or are there other rules of thumb that can be used? I'll be very grateful for any hints or tips on references. Also, if anyone knows if it is possible to use these fit statistics to test which distribution fits the data best, that would also be wonderful!

Thank you very much,