Related Experiment Videos
Smooth estimation of the survival function for interval censored data.
1Division of Biostatistics, School of Public Health, University of Minnesota, A460 Mayo Building (Box 303), Minneapolis, MN 55455-0378, USA. weip@biostat.umn.edu
Statistics in Medicine
|September 15, 2000
Summary
This study introduces and evaluates two smooth estimators for survival functions with interval censored data, showing they outperform the standard non-parametric maximum likelihood estimator (NPMLE) for smooth survival data. The integrated weighted difference (IWDB) test is highlighted as promising for comparing survival curves.
Area of Science:
- Biostatistics
- Survival Analysis
- Statistical Modeling
Background:
- Interval censored data are common in longitudinal studies, where event times are only known within intervals.
- The standard non-parametric maximum likelihood estimator (NPMLE) for survival functions produces step functions, which may not be efficient for smooth underlying survival distributions.
- Existing smooth estimators, including those based on NPMLE and logspline models, lack comprehensive finite sample performance evaluations.
Purpose of the Study:
- To evaluate the performance of two smooth estimators for survival functions with interval censored data.
- To compare the efficiency of smooth estimators against the NPMLE for smooth survival functions.
- To assess bootstrap-based test statistics for comparing two survival curves derived from interval censored data.
Main Methods:
- Simulation studies were conducted to compare the performance of NPMLE with two smooth estimators (one based on NPMLE, another using logspline density models).
- Two test statistics, the Kolmogorov-Smirnov test and the integrated weighted difference (IWDB) test, were investigated for comparing survival curves using bootstrap methods.
- The proposed methods were applied to reanalyze the Breast Cosmesis Study data.
Main Results:
- Simulation results demonstrate that both smooth estimators offer improvements over the NPMLE when the true survival function is smooth.
- The IWDB test showed particular promise for comparing survival functions, especially in cases of stochastic ordering without proportional hazards.
- The application to the Breast Cosmesis Study data illustrated the practical utility of the developed methods.
Conclusions:
- Smooth estimators are more efficient than NPMLE for interval censored data when survival functions are smooth.
- The IWDB test, utilizing bootstrap, is a valuable tool for comparing survival curves, particularly under stochastic ordering.
- The study provides practical guidance and validated methods for analyzing interval censored survival data.