I ran a Cox-Regression on Prepayment analysis. Right now, I'm using a stratified Cox-Regression with time dependent variables. The focus of my work lays NOT on prediction but rather on differentiating among risk groups.

The problem is that my p-values are extreme low, while my (pseudo) R-square is extremely low, around 0.001 (max 0.225) . Concordance is around 0.50

What do these results tell me? I heard different comments on the R-Square measure. Some say that it is complete meaningless since it does not handle censored data correctly (same for concordance)

Can I simply ignore this measures? Or is there one way to push R-Square?

  • $\begingroup$ For an extendend discussion, one should read the post from Ben Kuhn at in: stats.stackexchange.com/questions/116540/… He proposes a "Harrrels C" index, tough I'm not quite sure if this is actually practically appliable. $\endgroup$
    – Kosta S.
    Oct 17, 2017 at 14:01


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.