报告题目：A consistent specification test for expectile models

主讲人：宋晓军副教授（北京大学）

时间：2024年8月21日（周三）16:00 p.m.

地点：北院卓远楼305会议室

主办单位：统计与数学学院

摘要：In many economic applications, interest goes beyond the mean of the conditional distribution of the response given the covariates. Quantile regression (Koenker and Bassett, 1978) allows the researcher to explore a continuous range of conditional quantile functions, Drawbacks of quantile regression: non-differentiable loss function, inefficiency for Gaussian-like error distributions, and difficulty calculating acovariance matrix. Expectile regression (Newey and Powell, 1987) has emerged as a powerful alternative to quantile regression in a wide range of applications. Advantages over quantiles: easier to compute; covariance matrix of an expectile estimator does not involve the density function. We develop omnibus specification tests of parametric expectile models over a continuum of expectile levels. The test has nontrivial power to a sequence of local alternatives approaching the null at the rate n-1/2, A simple multiplier bootstrap procedure is proposed to implement the test.

主讲人简介：

宋晓军，男，北京大学光华管理学院商务统计与经济计量系副教授，博士生导师，西班牙马德里卡洛斯三世大学经济学博士。主要研究兴趣是理论计量经济学，包括非参数/半参数方法，假设检验和自助法，以及计量经济学的应用等。论文发表在Annals of Applied Statistics，Biometrics，Econometric Theory，Journal of Applied Econometrics，Journal of Business & Economic Statistics和Journal of Econometrics等国际期刊。主持和参加自然科学基金面上项目和重点项目等。自2020年1月起担任Economic Modelling副主编。