AN APPROACH OF INFORMATION CRITERIA FOR MODEL SELECTION IN STATE SPACE

Jean Pierre López Vargas¹; Paulo Battaglin¹; Gilmar Barreto¹

¹ Universidade Estadual de Campinas

doi:10.20906/CPS/CILAMCE2015-0797

Resumo

In this article, we show a form to evaluate mathematical models represented in state space through the Akaike Information Criterion (AIC). We use the maximum likelihood method for estimation the set of optimal parameters of the model and the recursive Kalman Filter algorithm for estimating the state of same model. Thus we estimate the statistical quality of models by the Criterion of Akaike. This article proposes to clarify and make it simpler to understand the AIC in state space, which will be shown through an example.

Palavras-chave: Akaike Information Criterion; Maximum likelihood method; Recursive Kalman Filter; Time series; State Space models