Seminar Details

Seminar Details


Monday

May 9

3:30 pm

Binary Tree-Structured Statistical Methods and the Genetics of Hypertension

Richard Olshen

Seminar

Stanford University - Department of Electrical Engineering & Statistics

The talk is about binary tree-structured approaches to classification, clustering, and regression. A new approach to classification, FlexTree, will be discussed, as will be its application to separating hypertensive and hypotensive individuals by information gathered at clinical visits and also regarding SNPs (single nucleotide polymorphisms). I will discuss successive two-means clustering (pruned, tree-structured vector quantization -- PTSVQ) and its application to HIV V3 loop region amino acids, though this topic will receive less mention than will FlexTree. Finally, I will discuss some mathematical difficulties that attend the study of binary tree-structured regression and will relate them to problems in the differentiation of integrals.