Using cmdstanr with loo on brms models?

Hi everyone,

When fitting a brms model with backend = “cmstanr”, I don’t seem to be able to perform loo cross-validation. I get the following error: “Error: Backend ‘rstan’ is required for this method.” As the cmdstanr backend provides such an efficient speed-up, I’d love to be able to use this backend with loo.

Is there a specific reason why this backend can’t be used with loo? Are there any workarounds, or is it in development? Thanks to anyone who knows anything!


Answers here:

1 Like