Premium
SOME SMALL–SAMPLE NON–PROPORTIONAL HAZARDS RESULIS FOR THE KAPLAN–MEIER ESTIMATOR
Author(s) -
Geurts J.H.J.
Publication year - 1985
Publication title -
statistica neerlandica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.52
H-Index - 39
eISSN - 1467-9574
pISSN - 0039-0402
DOI - 10.1111/j.1467-9574.1985.tb01120.x
Subject(s) - censoring (clinical trials) , estimator , mathematics , statistics , proportional hazards model , survival analysis , hazard ratio , kaplan–meier estimator , mean squared error , econometrics , confidence interval
CHEN, HOLLANDER and LANGBERG (1982) presented exact small–sample results for the KAPLAN–MEIER (1958) estimator under the proportional hazards model. We present simulation results for the same estimator for small samples in a number of cases where the hazard rates are not proportional. These suggest that the proportional hazards results, especially with regard to the Mean Square Error, are typical for a fairly large variety of “non–proportionalities” that may be encountered in practice. Thus the MSE of the KAPLAN–MEIER estimator appears to be rather robust against deviations from proportionality of the survival and censoring hazard rates. We illustrate the practical importance of this property by two examples.