Testing procedures based on the estimation of the error
distribution in nonparametric regression models have been proposed in
the recent literature. In the first part of the talk, we will revise the
general idea of those testing procedures and present some examples
(tests for the parametric form of the regression function, comparison of
regression curves, tests about the equality of error distributions, and
tests for the parametric form of the variance function). In the second
part of the talk we will present some extensions and new perspectives of
this kind of tests. Firstly, we will show how censored data can be
incorporated to these models. Secondly, we will briefly present an
estimator of the error distribution with multiple covariates, and show
how it can be employed to test for additivity. Finally, we will discuss a
test for a multiplicative structure in a dependent data setup. |