0

In my understanding, cox PH model can only give estimates of coeffs of regressors and then interpretations can be drawn.

Since it does not assume a base hazard function, it cannot give an estimated survival curve S(t) like a parametric method does, right?

However, when I do cox in SAS, the procedure PHREG does produce the predicted S(t).

  1. Do we have to use a full-parametric model to predict S(t).

  2. Can PH model be modified so it could predict S(t). Is SAS using away to estimate the base hazard line so a S(t) is given by PROC PHREG.

Thanks guys.

John
  • 263
  • 2
  • 12
  • 3
    http://stats.stackexchange.com/questions/36015/prediction-in-cox-regression/36077#36077 – ocram Apr 24 '15 at 08:29
  • 2
    You can obtain an estimate of S(t) *without* assuming a parametric form for the base hazard; one can obtain a nonparametric estimate instead. In the case of Cox regression, the set of hazard functions could then be called semi-parametric. – Glen_b Apr 24 '15 at 10:20

0 Answers0