Delgado, M.A. and J. Mora
Econometrica – 63.6 (1995), 1477-1484

Paraules clau:: nonparametric regression, semiparametric inference, discrete regressors, partially linear model

Resum: This paper is concerned with nonparametric and semiparametric inference in regression models where regressors are not continuous. We show that pointwise root-n-consistency and global consistency of regression curve estimates are achieved without employing any smoothing, even for discrete regressors with unbounded support. We discuss how these results can be applied to the estimation of semiparametric models; we consider in detail the partially linear model. The relationship between the nonsmoothing estimate and other well-known nonparametric techniques is also analyzed.