Copy the page URI to the clipboard
Goodwin, Jennifer Ann Grace
(1975).
DOI: https://doi.org/10.21954/ou.ro.0000fca3
Abstract
The question of how to compare survival between two or more groups is considered mainly with a view to applications in medical studies. Emphasis is laid on finding methods of comparison which are simple to perform, and on the presentation of the results in a form which may readily be understood by the layman.
The main nonparametric methods of estimating a survival distribution are described, and the connections between them pointed out. Parametric methods considered are those based on the exponential and Weibull distributions. The comparison of two groups is considered in detail, particular attention being paid to simple graphical methods, and to the clarification of some points arising from the recent papers by Gox (1972) and Pete and Peto (1972).
Where account is to be taken of concomitant variables, the important regression models available are those based on the exponential distribution and Cox's model, in which the form of the underlying distribution remains arbitrary. These are presented as extensions ofthe two-sample models already considered. Where there are many possible covariates, computation becomes heavy and there is a danger that the results may be difficult to interpret. It is suggested that a preliminary analysis of the data in strata by simple two-sample methods be made, as an aid in deciding which covariates to include in the model; a subsequent regression analysis might then be presented as a refinement of this. Such a preliminary analysis is demonstrated on a body of data from a clinical trial.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)
