A0584
Title: Treatment effects for Interval valued data: A copula-based analysis
Authors: Richard Moussa - Ecole Nationale Superieure de Statistique et d\'Economie Appliquee (Cote d'Ivoire) [presenting]
Abstract: The estimation of the average treatment effect in a case where the treatment variable is endogenous and the outcome variable is interval valued is considered. A non parametric approach based on copula is used to overcome the limitation related to the joint normal distribution assumption. The main results obtained considering departure from joint normality are the following. Firsthand, selection procedure has low performance on small sample and the performance increases with sample size. Secondhand, tests for switching regression model have good performances using the copula based approaches for both small and large sample sizes. Thirdhand, the mean square error and the estimated standard errors are lower using copula based approaches than in parametric model.
