Zongwu Cai

Department of Mathematics

University of North Carolina - Charlotte


Nonlinear Markov Regression Models with Application to Longitudinal Data

In this talk, first I will review Markov regression models in the literature. Secondly, I propose a local quasi-likelihood approach to regression analysis with discrete longitudinal data. In particular, I consider a class of nonlinear time series models -functional-coefficient models, which extend the Markov regression models for time series proposed by Zeger and Qaqish (1988, Biometrics). The method I used to estimate the coefficient functions is a local linear fitting technique and the asymptotic property of the estimator is studied so that a consistent estimator of the asymptotic variance can be derived. A nonparametric version of Akaike information criterion is proposed to select the optimal bandwidth and the best model. A simulated example and a real dataset for respiratory disease are reported to illustrate the methodology.


Back to Colloquium Series