Title:  Longitudinal Analysis with Nonparametric Varying-Coefficient Models and 
        Time-Dependent Covariates

Name: Colin Wu

Affiliation: National Heart Lung and Blood Institute

Abstract: An important objective in longitudinal analysis is to evaluate the effects of 
          covariates, either time-invariant or time-dependent, on the time-dependent 
          response variables. Because of the special modeling structures, the time-varying 
          coefficient models enjoy the advantage of being both flexible and mathematically 
          tractable. In this talk, we propose a class of time-varying coefficient models 
          with centered or standardized (both centered and scaled) covariates, discuss the 
          practical benefits of covariate transformation, and develop a comprehensive set 
          of nonparametric estimation and confidence procedures based on these models. We 
          demonstrate the practical properties of our methods through an epidemiological 
          fetal growth study and a simulation study. Our results suggest that covariate 
          centering or standardization can lead to statistically superior estimators or 
          inferences and more desirable biological interpretations.   

(This is a joint work with Kai F. Yu and Vivian W.S. Yuan)