统计与管理学院2015年学术报告第1期

【主  题】 Full imputation smooth distance and related estimation for conditional estimating equations with missing data

【报告人】 林路   

 山东大学

【时  间】 2015年1月5日（星期一）09:30－10:30

【地  点】 上海财经大学统计与管理学院大楼1208室

【语  言】 英文

【摘  要】 In many areas of statistics and econometrics, there exist the models that are defined in terms of conditional estimating equations. When the data in the equations are incomplete, most of the existing techniques involve non-parametric estimation. These methods are then faced with the difficulty of selecting smoothing parameter, and moreover, are inefficient in estimation and computation for the case of multi-dimensional variables. To tackle these issues, we in this paper define a full imputation smooth distance for conditional estimating equations with missing data and then suggest a smooth minimum distance estimation for the parameters in the estimation equations. It is somewhat surprising that although a multivariate kernel estimator is used during intermediate procedure, the resulting parameter estimator is always $\sqrt n$-consistent and asymptotically normal for a fixed, non-vanishing bandwidth, as well as a vanishing one. As such, the new method is particularly useful when the involved variables are multi-dimensional.

【邀请人】 朱利平