How to compare 2 ORs from different regression models.

#1
I created a logistic regression model to predict an outcome of depression vs. no depression. There are several variables in the model including gender. I then stratified the population and created 2 logistic regression models (one for women and one for men) because of interactions between gender and another variable.

Would I be able to compare the ORs from the different models? For example,
Female model: college education OR= 3.0 (2.6-3.7)
Male model: college education OR= 1.5 (1.2-2.4)

Can I say that college education increases the odds of depression more so (or 2x as much) in women than in men? I was told you can only compare ORs from within the same model (using CIs) How would I compare these 2 ORs? A t-test?

Thanks!
 

CowboyBear

Super Moderator
#2
I don't see why you can't compare them as is. You don't need to do another test - it's obvious from the confidence intervals that there is evidence that the effects differ.

Why did you choose to estimate two models instead of just specifying interactions as part of your model though?