I think that there is a rough agreement in the original thread that when you care about missing responses, you can’t do much better than just infer the missing values from the model. If there is a reason, you believe this is not appropriate for your case, could you explain why do you think so?
Quoting the relevant excerpts:
I recommend this, too, unless Y is multivariate and only some values for each row are missing. It can still be possible that the likelihood factorizes and you can just drop likelihood terms for the missing cases.
Maybe I am missing something (and maybe you resolved your issue in the meantime), but what would be wrong about fitting only the data without missingness and using posterior_predict(fit, data = orig_data %>% filter(is.na(Y), allow_new_levels = TRUE)
to estimate the uncertainty you have about the missing Y
values? This just takes the fitted uncertainty in your factor levels and draws the coefficient for the combinations of predictors not seen in the non-missing data using this uncertainty. This needs to assume that the unobserved combinations are in some sense “from the same population” as the observed ones, so it won’t help you if there is systematic bias in the unobserved. But it might be a good start…
Best of luck with your model!
3 Likes