I am running an Cox regression with my data. In my final model I had four potential predictors and did first a forward (Wald method) Cox regression and then for validation of my model a backward (Wald method) cox regression. Unfortunately I got two different models. What can I do now? Is there any convention to decide which model is more stable?

Thanks a lot for your answers! :)


Welcome to the site.

What you have done is demonstrate one of the several problems with methods like forwards and backwards selection. These have been discussed here many times. This list has 62 posts about forward variable selection.

In short, I think you should start over and do something other than forward or backward; but what you should do depends on total N, the total number of potential independent variables, their relation to each other, the state of the literature on your subject, your research questions and so on.

| cite | improve this answer | |

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.