Estimando e Identificando na Classe de Modelos Lineares Hierárquicos Normais

Abstract

The objective of this paper is to examine problems of estimation, identification and prevision that ocenr in some normal hierarchical linear models. Firstly this discussion was for the first-order polynomial model both in its static and dynamic shape. Based on that analysis some conclusions were obtained and extensions were made from the mixture of distributions (Gamerman & Smith, 1996) in order to maintain the model identifiable with the possibility of a best estimation of the parameters. Considering this model as a generator element of the general class of Dynamic Normal Linear Model (DLMs), generalizations were made for polynomial models of order k and regression Models.