1
$\begingroup$

How should I compare my logistic regression models?
My study is about the association between diabetes (dia), and a genetic variation (GV) I'm adjusting for confounders such as gestational age (GA) and gender (GE).

I have these models:

the simpel, reduced model

dia ~ GV 

half complex, intermediate model

dia ~GV + GA 

The full complex model

dia ~GV + GA + GE 

All three models are testet from the same dataset. But I don't know how to present my models in my project. Can I compare the reduced model and the full model? Or should the "half" model be included in the project as well? I'm only interested in the coefficient for GV. So is it important to include all models? I'm asking because that would be a lot of forest plots to show, and I hope that it is okay to compare only the reduced and the full model. But if that does not make sense, then I will include all three. Hope you can help and clarify things for me Thanks in advance!

$\endgroup$

1 Answer 1

1
$\begingroup$

No presentation that you are considering would violate any statistical rules that I know of. It's a question of how much they would show about your results and what you found. You also might not need to show forest plots for all three.

You could

  • present just the complex model
  • present the simple and the complex one
  • present all three

And the presentation could be done in a table.

But, if this is a project for a journal or a school, then they may have their own requirements.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.