Classical parametric estimation methods applied to nonlinear regression and limited-dependent-variable models are very sensitive to misspecification and data errors. On the other hand, semiparametric and nonparametric methods, which are not restricted by parametric assumptions, require more data...