Simulated maximum likelihood estimation of multivariate mixed‐Poisson regression models, with application

This paper proposes and implements simulated maximum likelihood estimation of bivariate count models with unrestricted correlation pattern of unobserved eterogeneity. The implementation incorporates both antithetic acceleration and adjustment for first‐order simulation bias. Both the Monte Carlo simulation evidence and an application to health utilization data confirm that the method is reliable and feasible in data analysis.