|
Activity Number:
|
578
- Survival Analysis and Semiparametic and Nonparametric Models
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 2, 2017 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #324817
|
View Presentation
|
|
Title:
|
Constructing Trees with Censored Survival Data
|
|
Author(s):
|
Madan Kundu*
|
|
Companies:
|
Novartis Oncology Pharmaceuticals - Florham Park, NJ
|
|
Keywords:
|
survival trees ;
instability test ;
survival data ;
censored data ;
score process ;
Brownian Bridge
|
|
Abstract:
|
Clinical researchers are often interested to define interpretable prognostic classification rules for censored survival data both for understanding the prognostic structure of data and for designing future clinical trials. Survival trees (i.e., tree with censored survival data) based on recursive partitioning could be very useful in such cases. Most of the currently available survival tree construction techniques are not based on a formal test of significance. A formal test of statistical significance of this heterogeneity would be useful to guard against spurious findings. We have proposed SurvCART algorithm under conditional inference framework that selects covariate for partitioning via formal parameter instability test and then finds the optimal split for the selected partitioning variable via maximizing the dissimilarity in survival distributions. SurvCART algorithm is developed to identify subgroups with either varying survival distribution or varying censoring distribution. The operating characteristics of parameter instability test and comparative assessment of SurvCART algorithm are studied via simulation. Finally, SurvCART algorithm is applied to a real data setting.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2017 program
|
