FRIDAY, November 22, 2002
Time: 1:30 - 2:30 PM
Constant Hall Room 1037

Ph.D. Pre-defense
Title: Estimation of Parameters in Replicated Time Series Regression Models

Genming Shi
Department of Mathematics & Statistics, Old Dominion University

The time series regression model has been studied by a lot of authors, but the case with replicated measurements has seldom been studied. In this thesis, I discuss several parameter estimating methods regarding this model with ARMA (p, q) correlation structure, such as maximum likelihood method, moment method, and quasi-least squares method. The asymptotic comparisons are made by fixing the number of repeated measurements, but letting the number of replications go to infinity. Numerical results are presented to illustrate that the quasi-least squares estimates are better than moment estimates, and good competitors to MLEs and even better for some cases.