Application and comparison of time series models to AIDS data

by 1967- Haileyesus, Tadesse

A class of space time bilinear (STBL) models have been proposed by Dai and Billard (1998). In this paper we have applied these newly developed STBL models, existing space time autoregressive and moving average (STARMA) models, and the standard nonspatial linear autoregressive and moving average (ARMA) models to a set of reported AIDS data in U.S. and compared the results. The STBL models were found to be the best fit to our AIDS data; a discussion of the findings is given. Index words: Space Time Bilinear Model, Bilinear Time Series, Maximum Likelihood Estimation, Forecasting, Weighting with Probability Proportional to Size, STBL, STARMA, ARMA, HIV/AIDS Application and Comparison of Time Series Models to AIDS Data by Tadesse Haileyesus A Thesis Submitted to the Graduate Faculty of The University of Georgia in Partial Fulfillment of the Requirements for the Degree Master of Science Athens, Georgia 2002 c? 2002 Tadesse Haileyesus All Rights Reserved Application and Comparison of Time Series Models to AIDS Data by Tadesse Haileyesus Approved: Major Professor: Lynne Billard Committee: T. N. Sriram Nancy Lyons Electronic Version Approved: Gordhan L. Patel Dean of the Graduate School The University of Georgia May 2002 To my parents, Tsegemariam and Haileyesus, to my beloved wife, Saba, and to my children, Adam, Omega and Nathan. iv