Thank you! That’s very helpful. I’ll have to read more about bridgesampling (which I’ve now been doing). While I do that, I have a few more follow up questions, if you or anyone else has time:
-
For the fixed effects, I’ve set the prior for the replication data to be a normal with the mean of the mean of the posterior distribution of the same parameter from the original data, and the standard deviation set to the standard deviation. Does that sound ok?
-
Since I am working with a mixed model (with a bernoulli response distribution), I also need to set priors for the random effect variances and correlations. Do you have any recommendations? An inverse-gamma for the variances?
-
When I’m constructing the null model (setting one parameter of interest to 0 by removing it from the model), I am leaving the corresponding random slope for the effect by any grouping variable (in my case, the items in a repeated measures design) in the null model and only remove the fixed effect parameter.
Any thoughts on this would be appreciated.
Florian