A GENERALIZED F-MIXTURE MODEL FOR CURE RATE ESTIMATION

Cure rate estimation is an important issue in clinical trials for dise ases such as lymphoma and breast cancer and mixture models are the mai n statistical methods. In the last decade, mixture models under differ ent distributions, such as exponential, Weibull, log-normal and Gomper tz, have been discussed and used. However, these models involve strong er distributional assumptions than is desirable and inferences may not be robust to departures from these assumptions. In this paper, a mixt ure model is proposed using the generalized F distribution family. Alt hough this family is seldom used because of computational difficulties , it has the advantage of being very flexible and including many commo nly used distributions as special cases. The generalised F mixture mod el can relax the usual stronger distributional assumptions and allow t he analyst to uncover structure in the data that might otherwise have been missed. This is illustrated by fitting the model to data from lar ge-scale clinical trials with long follow-up of lymphoma patients. Com putational problems with the model and model selection methods are dis cussed. Comparison of maximum likelihood estimates with those obtained from mixture models under other distributions are included. (C) 1998 John Whey & Sons, Ltd.