In Bayesian Population Analysis, Kéry & Schaub use uniform priors for p & omega. But of course my model is different from theirs. Maybe I should put priors on p_bar_X & p_bar_X_combined, for example?
Edit: maybe I’m transforming the data wrongly before feeding it into Stan. Bc the model still produces bad estimates even after commenting out the partial pooling. Looks like it might be training on transformed arrays or something like that?
Edit II: I summed the input columns in generated quantities & all of them had the right values. So I guess something’s the matter with the model after all!