Comment on "Wang et al. (2005), Robust estimating functions and bias correction for longitudinal data analysis"

This note provides a discussion on the manuscript by Wang et al. (2005) who aim to robustify inference for longitudinal data analysis by replacing the ordinary generalized estimating function with an influence-bounded, possibly biased, version. To adjust for the bias of the ensuing robust estimator, the authors provide its analytic approximation by means of asymptotic expansions, and estimate it by plugging-in a nonrobust estimate of the parameter of interest. In this letter, we argue that the proposed bias-corrected estimator is, in fact, nonrobust.