Thank you @colcarroll for your reply and comments!
- Convergence is fine, i.e. no divergences and rhat virtually one.
- The dataset is large, 50K records in cluster 1 and 10K in cluster 2. What do you exactly mean with “it means you could add more features to the model”?
- The large sample size probably doesn’t warrant Bayes here, as you mention. But in terms of interpretability of results I think Bayes has advantages over e.g. a Wilcoxon rank-sum test (or Hypothesis testing in general), since it also gives you the magnitude of the difference. What are your thoughts on this?