A generalized partially linear framework for variance functions (Q1786908)

The authors are concerned with the so-called variance function generalized partially linear model (VFGPLM), a heteroscedastic regression model where both the mean function and the variance function are modeled in a partially linear manner. A two-step estimation procedure for the parameters of the linear part of the mean and the variance function, respectively, is proposed. The authors prove asymptotic normality and analyze (asymptotic) efficiency of the resulting estimators. Finally, the finite-sample behavior of the proposed estimators is illustrated by means of computer simulations and by an empirical example.