科研成果详情

Inferences for generalized linear mixed models are greatly hampered by the intractable integrated likelihood. In this paper numerical integration based on Quasi-Monte Carlo method is used to approximate the integral of the EM algorithm and then to fit the models. The proposed algorithm is computationally straightforward and easily implemented, and yields satisfactory estimates, compared to Monte Carlo approximation and Gauss-Hermite quadrature with the same computational cost.