Using Jeffreys' Non-Informative Prior Distribution in Bilinear Time Series Modeling With an Application on Ulu Serting Rainfall Data

Abstract

A study on Bayesian approach for modeling purposes started with its application on linear model. This can be extended to include linear time series model such as the autoregressive model, refer [1]. Bayesian analysis basically involves the determination of prior and posterior distributions, preferably those that fall under the conjugate families. The idea can be extended to nonlinear time series model such as the bilinear model. In this paper, we follow closely the method used by [2]. The prior distribution is the improper prior suggested by [3]. It is shown that the resulting posterior distribution is normal-gamma. It is a common belief that an environmetric data usually contain nonlinear characteristics. A numerical treatment of a local rainfall data will be presented.