Bayesian Selection Of Threshold Autoregressive Models PDF Download

This dissertation, "Bayes and Empirical Bayes Estimation for the Panel Threshold Autoregressive Model and Non-Gaussian Time Series" by Ka-yee, Liu, 廖家怡, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of the thesis entitled BAYES AND EMPIRICAL BAYES ESTIMATION FOR THE PANEL THRESHOLD AUTOREGRESSIVE MODEL AND NON-GAUSSIAN TIME SERIES Submitted by Liu Ka Yee for the degree of Master of Philosophy at The University of Hong Kong in January 2005 A panel of time series is a collection of time series that have similar character- istics.Paneltimeseriesanalysisreferstothepoolingofinformationinanumberof similar time series in order to improve the efficiency of statistical inference about the panel or the individual series. Panel time series contains more information about the data than a single series, thus can give more accurate estimations and predictions. While panel data analysis is widely used in the statistical literature, relatively little attention is paid on panel time series analysis. With the growing availability of data and computer power, panel time series will be an important tool in data analysis. Some authors studied the problem of estimation of panel autoregressive time series. Pooling of the series is one of the solutions but the uniqueness in eachof the series cannot retained. Bayesian approach is another solution. However, it usuallyimposestoomanyassumptionsonthepriordistributionwhichareusually not appropriate. Li and Hui considered an empirical Bayes procedure to estimate parameters of panel autoregressive time series model. On the other hand, Nandram and Petruccelli suggested an hierarchical Bayes approach. In the literature, little has been done on the analysis of panel non-linear or non-Gaussian time series. This thesis study attempts to fill this gap in the literature by considering estimation procedures for panel non-linear and non-Gaussian time series. The estimation procedures were unified under a Bayes-Empirical Bayes framework. Simulation results reveal that the three proposed Bayesian methods can im- prove the least squares estimates of the panel threshold autoregressive time series and the quasi-likelihood estimates of the panel non-Gaussian time series models. Their practicability was then demonstrated by analysis of real data. DOI: 10.5353/th_b3070616 Subjects: Bayesian statistical decision theory Time-series analysis