Sparse partial least-squares regression for high-throughput survival data analysis

Title
Sparse partial least-squares regression for high-throughput survival data analysis
Authors
이우주
Keywords
high-dimensional problem; partial least-squares; penalized likelihood; sparsity; survival analysis
Issue Date
2013
Publisher
STATISTICS IN MEDICINE
Series/Report no.
STATISTICS IN MEDICINE ; Vol32 no.30 Startpage 5340 Endpage 5352
Abstract
The partial least-square (PLS) method has been adapted to the Cox’s proportional hazards model f or analyzinghigh-dimensional survival data. But because the latent components constructed in PLS employ all predictorsregardless of their relevance, it is often d ifficult to interpret the results. In this paper, we propose a n ew for-mulation of sparse PLS (SPLS) procedure for survival data to allow simultaneous sparse variable selectionand dimension reduction. We develop a computing algorithm for SPLS by modifying an iteratively reweightedPLS algorithm and illustrate the method with the Swedish and the Netherlands Cancer Institute breast cancerdatasets. Through the numerical studies, we find that our SPLS method generally performs better than the stan-dard PLS and sparse Cox regr ession methods in variable selection and prediction.
URI
http://dspace.inha.ac.kr/handle/10505/33080
ISSN
0277-6715
Appears in Collections:
College of Natural Science(자연과학대학) > Statistics (통계학) > Local Access Journal Papers, Reports(통계학 논문, 보고서)

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse