Proper dispersion state space models for stochastic volatility

This paper introduces a fairly general class of state space models, based o n the notion of a proper dispersion model and on a suitable Markovian assum ption, which can be fruitfully employed for stochastic variance description . This class is obtained by translating to state space models a new mixing procedure, which generalizes the "studentization" technique of Jorgensen (1 997a). A preliminary investigation of the corresponding statistical propert ies is given and simple examples are presented.