Student Seminar Series - May 19, 2005
University of Minnesota
School of Statistics
College of Liberal Arts

Statistical Modeling of Multivariate Longitudinal Binary Data with a Latent Variable


Rong Yang


Thursday, May 19, 2005
3:00 PM, 127 Ford Hall
Minneapolis, East Bank Campus

Refreshments at 2:30 PM
300 Ford Hall


Abstract

 Multivariate longitudinal binary (MLB) data are commonly seen in applications. This thesis research focuses on developing new statistical modeling procedures for analyzing MLB data with a latent variable. A two-stage model is proposed for describing the time course of a latent variable which is believed to drive all the observable response variables, and for describing the relationship between the underlying variable and response variables. Based on the generalized estimating equations method, a blocked iterative algorithm is proposed for parameter estimation. This model is generalized in three different ways to accommodate both time-invariant and time-variant covariates. This research is motivated by a tantrum data set, which has eight longitudinal binary responses with a latent variable, denoting the anger intensity, and several covariates.