|
Activity Number:
|
178
- Recent Development on the Analysis of Time-to-Event Data
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Lifetime Data Science Section
|
|
Abstract #314047
|
|
|
Title:
|
On Consistent Variable Selection for Semiparameteric Mixture Cure Models
|
|
Author(s):
|
Yongzhao Shao* and Zhaoyin Zhu
|
|
Companies:
|
New York University School of Medicine and NYU School of Medicine
|
|
Keywords:
|
adaptive LASSO;
SCAD;
mixture cure models;
variable selection;
consistency;
oracle property
|
|
Abstract:
|
Semiparametric mixture cure models have been increasingly used for analyzing time-to-event data in many applications including cancer research where there are two latent groups of patients, those who could eventually experience events and those who become immune or cured after certain treatments. For many cancer patients, a large number of variables are routinely collected in many clinical and medical examinations including those with significant predictive or prognostic importance. The successful implementation and application of mixture cure models highly depend on the identification of important risk factors that affect the cure probability and/or the survival distributions among uncured subjects. However, there is a lack of rigorous justification for variable selection in this context due to the challenges related to unknown cure status and heavy data censoring. The aim of this paper is to establish validity and asymptotic optimal properties of the penalized likelihood based methods for variable selection in mixture cure models with adaptive lasso and other penality functions. The finite sample properties of several different types of penalties and tuning parameter selection criteria are compared in simulation studies. The performance of the variable selection methods is also illustrated using a cohort of melanoma patients at New York University Cancer Center.
|
Authors who are presenting talks have a * after their name.
