Model misspecified or only weakly predictive?

Can I just add a question here related to the priorsense package? Sorry but I feel interested in the package. I read that the package can work for models fitted both using brms and rstanarm packages. However, I just see how it works for brmsfit family, but not sure how to correctly define the arguments of the function powerscale_sensitivity(). @n-kall can you help with this? As I try with my model fitting using stan_glmer () with:
sensivity_base ← powerscale_sensitivity(
model_full_rstarnarm, variable = c(“Gender”, “Edu”))

Just taking two predictors as examples but it did not work and having this error:
Error in get_draws(x) : argument “get_draws” is missing, with no default


priorsense is currently not compatible with rstanarm models. It works with brms and cmdstanr / rstan models if the log prior and log likelihood evaluations are returned in the fit. This requires these variables to be specified in the Stan code (which is not done by rstanarm, and as it is using pre-compiled models, is not trivial to add).

I’d like to start working with priorsense as well.

Is including log_lik and log_prior in generated quantities sufficient?

Yes, that should be sufficient. The default variable name for the log prior is lprior, although this can be specified manually.