报告题目：Parametric Modal Regression with Contaminated Covariates

报告时间：2024.06.30   11:00-12:00

报  告 人 ：Prof. Weixing Song  Kansas State University

报告地点：理学院东北楼二楼报告厅(209)

Abstract：Modal regression provides a robust venue to describe how the response values of high frequency changes with the covariates. In this talk, we propose a parametric modal regression procedure based upon the Gamma distribution family, when covariates are contaminated with normal measurement error. Compared to existing methods, the proposed procedure has three notable merits. First, it can handle multiple covariates subject to normal measurement errors; second, it possesses a simpler bias corrected likelihood function in general and a tractable expression in some special cases, resulting in faster computation and more precise estimation if the data distribution is correctly specified; third, empirical evidence shows that the Gamma-distribution-based modal regression has certain robustness against misspecification of the distribution of the measurement error. Numerical studies are conducted to evaluate the finite sample performance of the proposed method.