交通大學、清華大學 統計學研究所 專題演講
題 目：A Two-Stage Dimension Reduction Method for Induced Response and Its
Applications
主講人：洪弘教授 (台灣大學流行病學與預防醫學研究所)
時 間：99年10月22日(星期五)上午10:40-11:30
(上午10:20-10:40茶會於交大統計所429室舉行)
地 點：交大綜合一館427室
Abstract
Researches nowadays usually encounter the curse of high-dimensionality where
many developed statistical models fail to overcome. To tackle this problem,
suffcient dimension reduction aims to search the central subspace (CS) in
which necessary information of covariates regarding response are contained.
Subsequent statistical analysis can then be implemented in a low-dimensional
space while preserving relevant information. Several studies are interested
in certain transformation of the response (the induced response) instead of
the original one. With dfferent transformations under consideration, the
corresponding CS may vary. When estimating this induced central subspace
(ICS), existing dimension reduction methods, however, may suffer the problem
of ineffciency. In this article, we propose a more effcient two-stage
estimation procedure to estimate the ICS. This approach is further extended
to the case of censored response. An application to combining multiple
biomarkers is also illustrated. Simulation studies and two data examples, the
Cardiac Arrhythmia Study and the Angiography Cohort Study, evidence the
usefulness of our method.
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 111.251.169.6