/blog/2021/11/10/ame-bayes-re-guide/

[andrewheiss]

A guide to correctly calculating posterior predictions and average marginal effects with multilievel Bayesian models | Andrew Heiss
https://www.andrewheiss.com/blog/2021/11/10/ame-bayes-re-guide/

[andrewheiss]

Originally posted by Joshua Wiley on 2021-12-07 10:06:30

If I'm not mistaken, the global grand mean does not ignore any
region-specific deviations unless the model has a linear link function
or you are calculating results on the link scale. It uses the "fixed"
effects part of the model only, but critically these fixed effect
estimates are themselves conditional on the random effects so you still
end up with a conditional prediction. See for example:
Hedeker, D., du Toit, S., Demirtas, H., & Gibbons, R. D. (2018). A note
on marginalization of regression parameters from mixed models of binary
outcomes. Biometrics, 74(1), 354--361.
https://doi.org/10.1111/bio...

[andrewheiss]

Originally posted by stefan on 2022-05-02 18:55:43

Yes, I also think this is important. The "grand mean predictions" here
appear to be conditional on a hypothetical "typical" or "average"
cluster

[andrewheiss]

Originally posted by Dr.MZ on 2022-06-17 08:54:21

I was wondering if for the continuous measure if there was some method
to measure and plot the points at which the mean difference between
autonomy is non-0, such as the frequentist Johnson-Neyman region of
significance? Maybe have it as a 0 and -.1 .1 ROPE in the figure but in
reverse, to show at what values of civil liberties the difference is
"significant".