Nonparametric Regression With A Parametric Start for Binary Response Models

In the last few decades the econometric literature has generated a number of parametric, nonparametric, and semiparametric methods for solving binary response models. These procedures differ in the relative strength of their assumptions, convergence rates, consistency, efficiency, and prediction. This paper considers an estimator that assumes a parametric start and then couples with a nonparametric correction factor. This model is compared with parametric estimators of the link function on a Monte Carlo study and an application using data from Spector and Masseo (1980) on effects of new teaching methods in economics on student grades. 

Ariga 2001.pdf (1.95 MB)

Author(s)

Publication Date