课程题目：Structured Nonparametric Methods for Longitudinal Data

课程安排：

主 讲 人： Colin O. Wu，Senior Mathematical Statistician at the National Heart, Lung and Blood Institute, National Institutes of Health.

课程摘要：

Regression models for the analysis of longitudinal data (or functional data) have been an intensive area of statistical research for the past four decades. Since the subjects in a longitudinal sample are independent and have repeated measurements over time, longitudinal data are ideally suited in biomedical studies for estimating the time-varying covariate effects, predicting the future outcomes and tracking the subjects’ health status and risk factors. A large number of estimation and inference procedures with longitudinal data have been developed for the parametric, semiparametric and nonparametric regression models. These models have two major focuses: (a) modeling the covariate effects on the conditional means of the outcome over time, and (b) modeling the covariate effects on the conditional distributions of the outcome over time. We survey in this course a number of important structured nonparametric longitudinal models, which have been extensively studied in recent years. As opposed to the unstructured nonparametric models, the structured nonparametric models are mathematically tractable and have better scientific interpretations. On the other hand, the structured nonparametric models also preserve the much needed model flexibility compared with the well-known parametric and semiparametric models. We present the local and global estimation and inference procedures for four major structured nonparametric models: (1) the time-varying coefficient models, (2) the shared-parameter models, (3) the nonparametric mixed-effects models, and (4) the time-varying transformation models.  We illustrate the applications of these estimation and inference procedures in four longitudinal epidemiology studies and clinical trials.