1
$\begingroup$

What are the advantages of using Cox proportional hazards model versus logistic GEE? Can one generate Kaplan-Meier remainder estimator curves using logistic GEE?

$\endgroup$

1 Answer 1

1
$\begingroup$

There are at least two advantages to Cox models for survival data:

  1. They deal with censoring, logistic regression doesn't.
  2. They allow survival time, rather than just survive/did not. This usually provides more power and is often a more relevant question.

You don't say what your variables are, but let's suppose they are "surgery vs. not" and "death". It's usually important to distinguish "died in one day" from "died in 10 years".

$\endgroup$
11
  • $\begingroup$ What about poisson gee which can tell me how many deaths at a particular time? $\endgroup$
    – user271077
    Feb 11, 2020 at 17:32
  • $\begingroup$ Poisson GEE does not deal with censoring. $\endgroup$
    – Peter Flom
    Feb 11, 2020 at 18:09
  • $\begingroup$ You mean if someone dies and then we don’t have anymore data with them and GEE can’t deal with I forgot MCAR or MAR? What about with the geeglm package where I can specify waves and if someone dies, I can just not record them in the data and the gee still runs perfectly? $\endgroup$
    – user271077
    Feb 11, 2020 at 18:19
  • $\begingroup$ No, I mean if someone DOESN'T die then you don't know how long they would have lived. $\endgroup$
    – Peter Flom
    Feb 11, 2020 at 18:21
  • 1
    $\begingroup$ If you want to "make up" and "guess" things, then you can do this. If you want to use statistical analysis, you should use the Cox model. $\endgroup$
    – Peter Flom
    Feb 11, 2020 at 18:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.