1.Département de Mathématiques,Université Mentouri de Constantine,Constantine,Algérie;2.LMPA Joseph Liouville, Centre Universitaire de la Mi-Voix,Université du Littoral-C?te d’Opale,Calais Cedex,France
Abstract:
We consider a general class of time series linear models where parameters switch according to a known fixed calendar. These parameters are estimated by means of quasi-generalized least squares estimators. conditions for strong consistency and asymptotic normality are given. Applications to cyclical ARMA models with non constant periods are considered.