An efficient MCEM algorithm for fitting generalized linear mixed models for correlated binary data

摘要: Generalized linear mixed models have been widely used in the analysis of correlated binary data arisen many research areas. Maximum likelihood fitting these remains to be a challenge because complexity function. Current approaches are primarily either approximate or use sampling method find exact solution. The former results biased estimates, and latter uses Monte Carlo EM (MCEM) methods with Markov chain algorithm each E-step, leading problems convergence slow convergence. This paper develops new MCEM maximize for generalized probit-normal data. At utilizing inverse Bayes formula, we propose direct importance approach (i.e. weighted integration) numerically evaluate first- second-order moments truncated multivariate normal distribution, thus eliminating c...