# Thread: GLMM versus logistic regression if no random effects

1. ## GLMM versus logistic regression if no random effects

Dear all,

I am working with binary data and I was wondering:

in a model that does not include any random effect (only fixed effect), logistic regression and GLMM provide similar results?

Thanks for the assistance,
Nicola

2. ## Re: GLMM versus logistic regression if no random effects

Tell us more about data (cluster/repeat measures or multiple structural levels) ? Yes they can provide similar results, though some results or observation may not be independent, so you want to control for this covariance, but you are saying it is not significant correct?

3. ## Re: GLMM versus logistic regression if no random effects

I have data that are clustered so I will end using the GLMM to include the cluster factor as a random effect.

But this was more a "theoretical" question, I am not a statistician so I do not really understand the equations behind this two procedures (binary logistic regression and GLMM). But for what I understand, the main difference between the two models (for a binary outcome) is that in the GLMM you can include the random effect.

So, just from a calculation point of view I am wondering if GLMM and binary logistic regression gives the same results if no random effects are included.

Thanks!

4. ## Re: GLMM versus logistic regression if no random effects

Are you able to perform a comparison yourself? I believe the results will be the same in theme but may be slightly different based on the program and model equation.

5. ## Re: GLMM versus logistic regression if no random effects

I have actually tried it with SPSS right now. I attach the two outputs compared.
It seems that the Odds ratio (B coefficient) is the same (purple circles).
However while the 95% CI are not the same, so while it is significant in the logistic regression, it is not significant in the GLMM (red circles).

Do you know why the 95% CI are different?

6. ## Re: GLMM versus logistic regression if no random effects

Well to start with provide outputs of models with the same Covariates. The SEs are different which could be base on the formulae or because the 2nd model has more terms (which may partially explained a similar portion of the dependent variable).

Sorry - I now see they were both the same models, but one had the dummy codes displayed. Another difference is that the intercepts are not the same.

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts