Title: Panel data AR(1) time series models with multiple complete break points under a Bayesian framework
Authors: Jitendra Kumar - Central University of Rajasthan (India) [presenting]
Abstract: Economic time series are more prone to change the trend due to any of discrepancy such as political changes, policy reforms, import-export markets etc. These changes are attaching serious concern on time series modeling by various researchers working in this area. Recently, the change point time series modeling of real GDP and non-life insurance data has been studied. A generalization of the PAR(1) model is considered in which mean and error variance are shifted. Panel autoregressive model with multiple break points present in all parameters as autoregressive coefficient, mean and error variance, are dealt with. A Bayesian approach has been applied to estimate the parameters as well as getting the highest posterior confidence interval. A strong evidence has been observed for the Bayes estimator compared to the maximum likelihood estimator. Simulation experiments and an empirical application on SARRC associations GDP per capita time series are shown to illustrate the performances of the model. The developed model is also extended for temporary shift model.