GAMLj mixed model logistic regression

General help and assistance with jamovi. Bug reports can be made at our issues page: https://github.com/jamovi/jamovi/issues . (If you're unsure feel free to discuss it here)
Post Reply
iainjgallagher
Posts: 15
Joined: Sun Sep 24, 2017 10:51 am

GAMLj mixed model logistic regression

Post by iainjgallagher »

Hi

I wonder if someone could help me with the interpretation of a logistic regression mixed model.

I am trying to model response (either high or low) within control or intervention groups at two timepoints (before or after intervention or placebo). The omv file is attached. I'm using a mixed model just to account for paired observations (i.e. I only have a random intercept assuming the slope would be the same for everyone and the same across the groups).

My problem is in interpreting the Estimated Marginal Means i.e. the probability of answering low.

In the GAMLj output for example the probability of answering low in the control group is 1 (first Est Marginal Means table).

The actual numbers are ctrl:low = 14; not ctrl:low = 60. So P(low|ctrl) = 14/74 = 0.189

The df and CI's are also rather odd.

I'm rather new to mixed models so if someone could gently talk me through the interpretation (or more likely the mistake) here I can probably work out the rest myself.

Thanks.
Attachments
emot_probs_data_IJG.omv
(39.24 KiB) Downloaded 224 times
User avatar
mcfanda@gmail.com
Posts: 457
Joined: Thu Mar 23, 2017 9:24 pm

Re: GAMLj mixed model logistic regression

Post by mcfanda@gmail.com »

Hi,
your estimated probabilities are either almost 1 or almost 0. It is indeed a bit weird, but I've checked your models and I get the same results in R glmer() and emmeans(), so the results seems to be correct. On the side, I would not use "dummy" as the coding methods because with "dummy" you obtain simple effects rather than main effects. Use "simple" so you get the main effects that you probably need
Post Reply