Pseudo-maximum de vraisemblance: expériences de simulations dans le cadre d'un model de Poisson

This paper presents a study of the finite sample properties of the pseudo-maximum likelihood estimators in the case of a Poisson model. The simulations have been realized with control variates, a method which has considerably improved the accuracy of the results obtained. The negative binomial p. d. f. gives better estimators, for the pseudo-maximum likelihood method, than gamma or normal p. d. f. The quasi-generalized pseudo-maximum likelihood method really improves the results, whichever of those three p. d. f.'s is used.