2
$\begingroup$

I want to compare three groups on determinants: A group that has survived <5 years, a group that has 5-10 years of survival, and a group that has >10 years of survival. The question is whether certain determinants are different in the group with long survival i.e. What determinants are associated with survival of >10 years?

I've already established I can't use a Cox regression to answer this question (because it doesn't tell me specifically what characterizes those patients with >10 year survival, but rather what determinants are associated with events during follow-up).
Is ordinal logistic regression a good fit to answer this question? Or do you have other suggestions to answer this question?

I've also considered to do a subgroup Cox regression analysis of only those patients with >10 years survival and compare the hazard ratios with the Cox model from the whole cohort, but I'm afraid that will give me a problem with the statistical power (as there are only 900 patients in this group with 120 events compared to more than 4000 with 700 events in the whole cohort).

Improve this question
$\endgroup$
3
  • 2
    $\begingroup$ What you propose is often called discrete time survival analysis. I do not use Stata and requests for help with code are off-topic here anyway but I believe that a search for "discrete time survival analysis Stata" will prove very fruitful for you. $\endgroup$
    – mdewey
    Jul 14, 2017 at 12:52
  • $\begingroup$ @mdewey: I don't use Stata (I use R), but thanks for the answer anyway. $\endgroup$
    – Tami
    Jul 14, 2017 at 13:00
  • 1
    $\begingroup$ Sorry, misread the question. Substitute R for Stata may well help, there are many relevant R packages. $\endgroup$
    – mdewey
    Jul 14, 2017 at 14:20

1 Answer 1

Reset to default
1
$\begingroup$

I believe you are right and I don't see anything wrong with using ordinal regression as long as its assumptions like the parallel lines assumption are not violated.

That being said, Other models such as bayesian hierarchical models can be applied as well but I believe ordinal regression sounds plausible

Improve this answer
$\endgroup$

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.