2

I am trying to run a logistic regression with one class (class 0, n = 513) over-represented compared to the other one (class 1, n = 173).

So far, in order to correct for the biased intercept, I just resampled the class 0 in order to get n = 173, then ran a logistic regression, then stocked the estimates. Finally I calculated the median of all these coefficient.

The issue I have is that I am not so sure my method is correct, and I am not so sure how I could report z values and other stats usually obtained in summary of glm.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
  • How is the model performing? Misclassification error/ AUC – Sada93 Aug 08 '17 at 15:35
  • 1
    Your assumption is that half of the population is in class 0. I am pretty sure that is incorrect, and that your original estimate of the constant is as unbiased as possible, while your "correction" actually introduces a significant bias to your estimate. – Maarten Buis Aug 08 '17 at 16:45

0 Answers0