A Bayesian approach for identification of Additive Outlier in AR(p) model

Abstract

Time series is the way of data analysis and modelling in which present observation is retrieved based on past observations which is called ARIMA model in case of linear dependency. If series is contaminated by an outlier, then it affects both order and parameter(s). The present paper deals an autoregressive (AR) model with an additive outlier under Bayesian prospective. For identification of an outlier, posterior odds ratio has been derived under suitable prior assumptions. An empirical analysis and realization is carried out to get applicability of proposed testing methodology.

Keywords: Autoregressive model, posterior odds ratio, prior distribution.

AMS 2010 Mathematics Subject Classification: 62F03, 62F15, 62M10