【7月3日】Model Averaging in Linear Measurement Error Models

【大中小】【打印】【关闭窗口】来源：本站原创作者：统数院编辑：发布时间：2015-06-29

报告题目： Model Averaging in Linear Measurement Error Models

主讲人：张新雨博士（中国科学院）

时间：2015年7月3日16:10-17:10

地点：北院卓远楼305

主办单位：统计与数学学院

摘要：We develop model averaging estimation in the linear regression model where covariates are subject to measurement errors. The absence of the true covariates in this framework makes the calculation of the standard residual based loss function impossible. We take advantage of the explicit form of the parameter estimators and construct a new criterion. It has the property that it is asymptotically equivalent to the unknown model average estimator minimizing the loss function. When the true model is not included in the set of candidate models, the method achieves optimality in terms of minimizing the relative loss, while when the true model is included, the method estimates the model parameter with root-n rate. Numerical analysis in comparison with existing BIC and AIC model selection and model averaging methods strongly favors our new model averaging method.

张新雨博士简介：统计学博士（中科院），数学和系统科学研究所副研究员，主要研究领域为模型平均/选择，组合预测以及混合效应模型。在Annals of Statistics, Journal of the American Statistical Association, Biometrika, Journal of Econometrics等顶级统计学和计量经济学期刊发表过多篇论文，是中科院优秀博士学位论文奖（2011）获得者。