Doctor Yao Qiwei Talked on Problems of Variable Selection of High Dimensional Data

On April 6th，Statistics South Frontiers of Science lecture series had started in the mathematic building. Professor Yao Qiwei from London School of Economics and Political Science Department of Statistics was invited to give the first lecture.

With "Stepwise Searching in High-Dimensional Regression (high-dimensional data of variable selection)" as its theme, professor yao talked about how, in the case of linear regression, to choose a real variable factors, so as to achieve variable selection purposes when the number of p independent variables far beyond the number of samples n (p>> n) circumstances. . Professor Yao Qiwei and his collaborators improved the traditional BIC measure criterion, BICP and BICC two new criteria, so that in the p>> n the case to work (BICP) and good stability (BICC), and the variable selection method with the classical LASSO compared. Stochastic simulation data showed that both BIC-type criteria for outstanding performance. After the seminar, Professor Yao Qiwei patiently answered questions from the audience.

Professor Yao Qiwei primarily studies time series analysis, nonparametric regression, dimension reduction and factor modeling, space-time models, financial econometrics and so on. He has published more than 70 professional journal articles, and is the co-writer of the book, Nonlinear Time Series: Nonparametric and Parametric Methods withwell-known Chinese statistician, Professor Fan Jianqing.