Bandwidth selection for statistical matching and prediction

Abstract While there exist many bandwidth selectors for estimation, selection statistical matching and prediction has hardly been studied so far. We introduce a computationally attractive selector nonparametric out-of-sample problems like data matching, impact evaluation, scenario simulations or imputing missings. Even though the method is bootstrap based, we can derive closed expressions criterion function which avoids need of Monte Carlo approximations. study both, asymptotic finite sample performance. The derived consistency, convergence rate extensive simulation studies show successful operation selector. illustrated by applying it to real studying gender wage gap in Spain. Specifically, salary Spanish women predicted nonparametrically equation estimated men while conditioned on their own (i.e., women’s) characteristics. An important discrepancy between observed wages found, exhibiting serious gap.