Improving variance function estimation in semiparametric longitudinal data analysis

Abstract

We propose an efficient and robust method for variance function estimation in semiparametric longitudinal data analysis. The method utilizes a local log-linear approximation for the variance function and adopts a generalized estimating equation approach to account for within subject correlations. We show theoretically and empirically that our method outperforms estimators using working independence that ignores the correlations. © 2011 Statistical Society of Canada.