Seminar: Testing Multiple Hypothesis in Big Data Analysis

2015-06-24  Xiaodong Pan Hits:[]

Topic Testing Multiple Hypothesis in Big Data Analysis

Speaker: Dr. Zhigen Zhao

Abstract:

When testing one single hypothesis, it is given in every introduction statistics textbook that one should reject the null hypothesis if the p-value is less than or equal to alpha, a designated level reflecting one's willingness of tolerating the type I error. In modern application, it is a common practice for a scientist to test thousands, or even millions of hypothesis simultaneously. Can we still use the simple rule stated earlier?

In this talk, I will present the issue of the multiplicity when testing multiple hypothesis and the recent development on this area in the last two decades. I will also discuss my recent work on the construction of the optimal multiple testing procedures using the decision empirical Bayes approach. I will highlight a few applications with big data, arising from studies such as the genetics and social science.

Time:10:15--11:15, June 29, 2015

Room:X2511

Profile:Dr. Zhigen Zhao is an assistant professor at Department of Statistics, Temple University (USA). He got his Ph.D. in mathematics from Cornell University in 2009.

Pre:Seminar: Envelope Models and Methods Next:Seminar: Estimation of the Error Auto-Correlation Matrix in Semiparametric

Close

Baidu
sogou