2
$\begingroup$

I have a standard event history data set, which I used to fit a Cox proportional hazards model in Stata and R. The time variable is measured in days. There are no time-varying covariates on the RHS of the equation.

I have recovered the usual output (coefficients, hazard ratios, SEs, etc.) as well as the predicted baseline survival for each observation. What I need to do now is estimate the predicted number of events in the next 30 days. Any ideas on how to do this?

Thanks very much!

$\endgroup$
  • 1
    $\begingroup$ Welcome! Would be easier if you could clarify: do you mean you have calculated coefficients for something like the last 30 days? Also are you trying to extrapolate the data into the future for the current set of subjects or predicting for new subjects? $\endgroup$ – dardisco Nov 1 '13 at 21:25
  • 1
    $\begingroup$ Thanks very much for your quick response! Let me try and clarify. I have fitted a Cox PH model to the data and have estimated the impact of a bunch of time-invariant covariates on the log of the hazard ratio ln(h/h_0). From this, I can compute the baseline survival and cumulative hazard for each observation using Stata's basesurv basechaz options. What I am trying to do is make predictions for my current set of subjects. Specifically, I need to answer the question: of the X number of subjects that have not yet experienced the event, which proportion will experience it in the next 30 days. $\endgroup$ – user32197 Nov 1 '13 at 21:32
  • $\begingroup$ Better! Still need more clues... What have you tried? What is your expected output? Please edit original question rather than reply in order to make life easier for future viewers to understand... $\endgroup$ – dardisco Nov 30 '13 at 7:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.