Corrected group prognostic curves and summary statistics.

A method is presented for correcting distribution function estimators for distortions in the joint distribution of covariates in the subsamples receiving different treatments. This form of standardization is useful for depicting multivariate analyses and predicting population behavior. Furthermore, it also proves useful for comparing two trials which differ in patient mix and for adjusting estimates for know imbalances in patient characteristics. Standardization though important, is not often used in clinical trials because of the complexities introduced by multiple covariates. The proposed method overcomes these difficulties in a wide class of analyses.