A note on Korn and Simon's measure of explained variation for survival models

Abstract

Among many proposals to estimate proportion of explained variation for survival data, Korn and Simon's measure is the only one actually looking at the variation of time. While this should be an advantage, its usage, as measured by citations, suggests otherwise. We look at some possible reasons for this. We find one, serious, but easily corrected, deficiency of the measure, and another property, which is natural, but not very practical when there are censored data.