Robust Estimating Equations with Data-Dependent Tuning Parameters in Longitudinal Data Analysis

摘 要：
Longitudinal data arises when measurements are made repeatedly over time on a same set of subjects. The statistical analysis of such data is of prime importance in many fields, especially biological and health sciences; and requires special statistical methods due to the possible interdependence among measurements on a single subject. In order to make valid inferences, the interdependence needs be taken into consideration. A number of approaches have been developed for analyzing longitudinal data, including random-effects models, marginal models and conditional models. Being a multivariate extension of the least squares, the GEE approach does not possess robustness properties and may be dominated by a small proportion of contamination. A new method is required that gives the desired amount of robustness at the minimum loss of estimation efficiency. In this paper, we propose asymptotically efficient estimation procedures for analysis of longitudinal data by adopting data-dependent tuning parameters. A two-stage procedure is developed for choosing such data-dependent tuning parameters.
个人简介：
Professor Wang obtained his PhD in 1991 (University of Oxford) and worked for CSIRO (2005-2010). Before returning to Australia, Professor Wang worked for the National University of Singapore (2001-2005) and Harvard University as Associate Professor in biostatistics. He is a tenured professor in data science at QUT. Professor Wang has published more than 130 papers in international journals (with 2500+ citations and h-index 28). He has extensive experience in student supervision and

extensive international collaborative experience.