I am performing a logistic regression analysis after performing a stepwise regression. I would like to know why the odds ratios of the logistic regression cannot be higher than the stepwise betas.
Asked
Active
Viewed 24 times
0
-
2The OR can definitely be larger than the parameter coefficients, since the ORs are just the exponential of the coefficients. Please clarify. – Stephan Kolassa Jun 02 '22 at 12:28
-
Sorry @StephanKolassa I am new to this type of analysis and I don't have much knowledge on the subject. I did a first stepwise regression analysis and then using the model I got from the previous analysis I did a logistic regression analysis. However, my supervisor told me this: For wave 5 and 6 there are more OR than betas in the previous table. I don't quite understand what he means, whether the ORs cannot be greater than the betas or it doesn't have to. Thank you in advance – Adrián P.L. Jun 02 '22 at 12:32
-
2Hm. It can't be that the magnitude of the ORs is larger or smaller than the betas, per my comment. It sounds like your tables are not consistent in terms of the information they provide. Best to ask your supervisor what they mean by their comment. (Incidentally, if you are doing inferential statistics, recall that p values are invalid after stepwise model selection.) – Stephan Kolassa Jun 02 '22 at 12:54