Title: A simple and practical approach towards testing global restrictions on general functions
Authors: Valentin Patilea - CREST-Ensai (France) [presenting]
Jeffrey Racine - McMaster (Canada)
Abstract: A simple bootstrap procedure is proposed for inference on vectors or functions in a general context that involves estimation only under the alternative, while constraints are imposed via choice of a suitable transformation of the unconstrained estimate. The procedure is quite general and applies directly to functions or derivatives defined by separable and non-separable regression models. It can be used with parametric, semi and nonparametric estimators without modification. Potential applications include, but are not limited to, inequality inference on mean or quantile regression models where the bounds depend on the model's covariates, checking monotonicity, convexity, symmetry, homogeneity, for multivariate functions.