统计学系系列讲座之225期

时间：2016年5月10日（星期二）上午10:00-11:00

地点：史带楼301室

主持人：朱仲义 教授 复旦大学管理学院统计学系

主题：Consistent Estimation for Distribution-uncertainty Regressions via Cross-sample and Semiparametric Methodologies

主讲人：林路 教授（山东大学金融研究院）

简介：Motivating by the famous Ellsberg paradox, ambiguity (distribution-uncertainty) is quantitively and qualitatively important in behavior finance. We consider a type of distribution-uncertainty regressions that contains endogenous variable regression and semiparametric regression as its special cases. For such models, however, classical estimating function does involve infinitely many nuisance parameters caused by the uncertain distributions. Consequently, the parameters of interest cannot be consistently estimated and the corresponding prediction is imprecise, even aimless. In this paper, cross-sample and semiparametric techniques, together with a hidden-constant function, are proposed for dealing with the infinitely many nuisance parameters. The resultant estimating function only contains the parameters of interest, and the estimators of them are always consistent and normally distributed with standard convergence rate. Moreover, the newly proposed methodologies can avoid the use of instrumental variable or nonparametric estimation even if actually the model under study contains endogenous variables or nonparametric components. On the other hand, the methodologies for numerical computation are simple, and the corresponding computation procedures are somewhat similar to those for the distribution-certainty models. The main difference from the classical regression analysis is that the estimation efficiency is related to the level of distribution-uncertainty.

统计学系

2016-5-9