PEMILIHAN MODEL DAN ESTIMASI PARAMETER BAYESIAN HIRARKI UNTUK MODEL SUBSET ARMA MENGGUNAKAN ALGORITMA REVERSIBLE JUMP SIMULATED ANNEALING MCMC

Abstract

If the Subset ARMA models fitted to real data, then the true value of the order and model parameters are often unknown. The purpose of this paper is to find an estimator-estimator for the order and parameters of the model based on the data. In this paper, the identification of order and subset ARMA model parameter estimation is done in a hierarchical Bayesian framework. Within this framework, the order and model parameters are assumed distributed priors. All order information about the characteristics and parameters of the model then expressed in the posterior distribution. Determination of the probability of the order and parameters of the posterior models requires the integration of the resulting posterior distribution, is an operation which is very difficult to do analytically. Here the algorithm Reversible Jump MCMC Simulated Annealing developed to perform the necessary integration through simulated posterior distribution.