Small Sample Properties of Two Survival Function Estimators Based on Incomplete Data.
FLORIDA STATE UNIV TALLAHASSEE DEPT OF STATISTICS
Pagination or Media Count:
For estimating an underlying survival distribution, we consider two estimators based on a randomly right-censored sample The traditional Product Limit Estimator PLE, introduced by Kaplan and Meier 1958, and a competitor, the Piecewise Exponential Estimator PEXE, introduced by Kitchin, Langberg, and Proschan 1980a. Under a proportional hazards model we present formulas for the mean and variance of the PLE and find upper and lower bounds for mean and variance of the PEXE. These new expressions for finite sample sizes are compared with known asymptotic results. Author
- Statistics and Probability