Varying-coefficient single-index model for longitudinal data

In this paper we consider a general class of varying coefficient single-index models for longitudinal data. This class of models provides a tool for simultaneous dimension reduction and the exploration of dynamic patterns. We develop an estimation procedure using Cholesky decomposition, local linear and backfitting technique. Asymptotic normality for the proposed estimators of varying-coefficient functions, link function and parameters will be established. Monte Carlo simulation studies show excellent finite-sample performance. We illustrate our methods with a real data example.