This might be a really naiive question but I couldn't find an answer anyway. Say we have two populations, test and training. We apply a logistic model of interest to the training data and get a bunch of coefficients and CIs, then we do the same with the test data. How can we determine whether the coefficients are significantly different between the two instances?

Is it some kind of Z-test?

  • 1
    $\begingroup$ Why do you want to do this? I do not see a theoretical foundation for such a test. Why can't you just pool the data, add interaction terms with "test" or "training" indicators and test their significance? I guess this is not your intention but it is the most classical approach $\endgroup$ – Knarpie Dec 21 '17 at 15:04
  • $\begingroup$ Thanks for your response. It's because I don't actually have access to the underlying data, just the two model summaries. $\endgroup$ – danielsbrewer Jan 8 '18 at 16:24

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.