Integer-valued models miming classical econometric models
Résumé
Econometric time series model can be defined to describe integer-valued observations. We present an integer-valued GARCH type model and two bilinear type models. The latter one, which can be useful in a context of inventory management or epidemiology, is based on a thinning operator allowing for negative values. We establish the existence of such a processes. The estimation of the parameters is tackled with the help of different methods: conditional likelihood and quasi-maximum likelihood. We give some asymptotic results for these estimators. Numerical examples and applications from social medicine are presented.