Department of Mathematical Sciences
|DATE:||Thursday, November 6, 2014|
|TIME:||1:15pm to 2:15pm|
|SPEAKER:||Qiyi Lu (Binghamton University)|
|TITLE:||Significance analysis for high-dimensional low-sample size data|
In this presentation I will first review a paper proposed by Wei et al in 2013, which studies a statistical test for equality of distributions or, a weaker hypothesis, equality of means of two high-dimensional low-sample size data. A three-step non-parametric algorithm, Direction-Projection-Permutation (DiProPerm), will be introduced and we will discuss the impact of choosing different projection directions and test statistics for different testing purposes. Second section of this talk will be some potential extensions of the method to multi-class cases.