Kernel Smoothing for Longitudinal Data and Semi-parametric Efficient Estimators
主 题: Kernel Smoothing for Longitudinal Data and Semi-parametric Efficient Estimators
报告人: Lu Wang (Department of biostatistics，Harvard University)
时 间: 2006-12-27 下午 1:00 - 2:00
地 点: 理科一号楼 1303
Longitudinal data analysis is challenged by the fact that observations from the same subject are likely to be correlated. In this situation, conventional profile-kernel method fails to yield a semi-parametric efficient estimator for the parametric coefficients in a semi-parametric model. Wang, Carroll and Lin proposed an iterative kernel generalized estimating equation (GEE) estimator, which accounts for within-cluster correlation and is more efficient. We recently derived the semi-parametric efficient score function and the semi-parametric information bound in general scenarios under the semi-parametric conditional mean model and showed that the iterative profile-kernel GEE estimator is actually semi-parametric efficient. In this talk, we will present different kernel smoothing techniques and discuss their performance when analyzing longitudinal or clustered data. We will also review the techniques for investigating semi-parametric efficiency and discuss the asymptotic properties for the iterative profile-kernel GEE estimator.
kernel smoothing; longitudinal/clustered data; generalized estimating equation; iterative kernel GEE estimator; semi-parametric efficient score; semi-parametric information bound.