|
Activity Number:
|
103
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #305428 |
|
Title:
|
The LASSO Method for Variable Selection for Right-Censored Data
|
|
Author(s):
|
Lili Yu*+ and Dennis K. Pearl
|
|
Companies:
|
The Ohio State University and The Ohio State University
|
|
Address:
|
Trumbull Court, Columbus, 43210,
|
|
Keywords:
|
sieve likelihood ; LASSO ; model selection ; right censored data
|
|
Abstract:
|
Tibshirani proposed a variation of the "lasso" method that was to minimize the log partial likelihood subject to the sum of the absolute values of the parameters being bounded by a constant in Cox's proportional hazards model. Due to the nature of this constraint, it shrinks coefficients and produces coefficients that are exactly zero. We apply this method to a class of semiparametric models (linear transformation models) in which the response variable is right-censored and the error is symmetric at zero but its distribution is unknown. We propose to use sieve-likelihood method to calculate the log likelihood and the parameters simultaneously. Simulations indicate using sieve-likelihood to calculate the lasso criteria in this setting can pick approximately the correct number of zero coefficients.
|
