Thanks @nhuurre! However, the bad loo values result from the same model with “good” loo values, the only difference is using the extra code to extract those coefficients.
In other words, when I do not attempt to extract the coefficients from the model, I get fine loo values. When I do use the extra code to extract the coefficients in the same model with otherwise the same code, I get “very bad” loo values. Since the loo is estimated using the log lik, I have no idea as to why this would occur, as the log lik isn’t changing. The only change is just extracting and manipulating the coefficient values in the transformed parameters block…
Any ideas?
Thanks!