4

According to my modest understanding, the Cox model is all about estimating hazard ratios, relative measures of risk. However, it is possible (in R using e.g. the packages pec or peperr) to estimate absolute risks by a Cox model.

How is this done mathematically?

Richard Hardy
  • 54,375
  • 10
  • 95
  • 219
miura
  • 3,364
  • 3
  • 21
  • 27
  • 3
    The absolute risk at time t given a covariate vector x_i can be estimated as exp(-lambda(t)exp(x_i'beta), where lambda(t) is the Breslow estimator for the cumulative hazard. I don't know how this is derived though. – mogron Jul 31 '12 at 08:59

0 Answers0