Modelo de mistura com dependência Markoviana de primeira ordem

Abstract

We present the mixture model with first order dependence, MMM(1). This model corresponds to a redefinition of the hidden Markov model (HMM) where a non observable variable is used to control the mixture. The usual mixture model is a particular case of the MMM(1). The proposed redefinition makes easier the application of usual estimation tools as the EM algorithm. We present the maximum likelihood and Bayesian estimators for the normal and binomial cases of the MMM(1) and usual mixture models. Simulation studies show the functionality of the proposed models and their estimators. And finally we present an application to a real data set for the binomial case.