Hierarchically Penalized Cox Regression for Censored Data
with Grouped Predictors and Its Oracle Property
Sijian Wang
Department of Biostatistics
University of Michigan
Faculty Candidate
Friday, February 29, 2008
12:00 pm - 1:00
5275 MSC
| ABSTRACT |
In many biological and other scientific applications, predictors are grouped. For example, in genome study, the genomic data can be divided into biological meaningful groups such as genes belonging to the same pathways. When studying the dependence of survival outcome on these grouped predictors, it is desirable to select important predictors at both the group level and the within-group level. We develop a new variable selection method that utilizes the grouping structure for the Cox proportional hazard model. Our new method not only effectively removes unimportant groups, but also maintains the flexibility of selecting predictors within the identified groups. We also show that the new method offers the potential for achieving the asymptotic oracle property.
|
Return to seminar list
