In this paper, we present a generalized estimating equations based estimation
approach and a variable selection procedure for single-index models when the
observed data are clustered. Unlike the case of independent observations,
bias-correction is necessary when general working correlation matrices are used
in the estimating equations.