Title: Testing constancy of structural parameters in the direction of random alternatives
Authors: Miguel A Delgado - Universidad Carlos III de Madrid (Spain) [presenting]
Abstract: The problem of testing the hypothesis of constancy of structural parameters versus the alternative that they are an unknown function of an observed variable is considered. This includes testing that marginal effects are constant in regression models versus non-parametric interactive effects, and specification testing of panel data models. Test statistics depend on partial sums of concomitants involving residuals, which resemble classical parameter stability tests, where parameters vary with time under the alternative hypothesis. We also consider the problem of testing that a subset of coefficients is constant, which includes testing the specification of partially linear regressions versus random slope coefficients. The test is implemented using bootstrap approximations of the distribution of the statistic under the null hypothesis. We study the finite performance of the test by means of Monte Carlo experiments.