A Functional Logistic Regression Model for Longitudinal Data

Abstract

Functional linear models are useful in modeling longitudinal data with continuous outcomes and have been extensively studied in recent years. In many medical and health cohort studies, however, one often has to deal with dichotomous outcomes. In this talk, I will discuss a functional logistic regression model for analysis of binary longitudinal data. A two-step method is proposed to estimate the coefficient functions. Their asymptotic properties are studied. The developed method allows the covariate effects to vary over time. It can also be used to assess goodness-of-fit of parametric models. The procedure can be conveniently implemented using existing statistical packages such as SAS and Splus. I will illustrate the method using both real data and simulated data. A simulation study will also be presented to compare its performance with the GEE method.