Pairwise Likelihood for Generalized Linear Models with Crossed Random Effects

摘要

Parameter estimation in Generalized Linear Models with crossed random effects is made difficult by the high-dimensional integrals required to obtain the full distribution of the response. We propose inference based on the pair-wise likelihood, which only requires the computation of bivariate distributions. The estimators based on the pairwise likelihood are generally consistent, and the efficiency loss with respect to maximum likelihood estimation is usually not substantial. The method is applied to the famous salamander mating data.