Generating LOO-CV predictions from stacked predictive distributions

Interfaces brms
,
1

I’ve fit a series of models and would next like to combine them via stacking and then evaluate the LOO-CV-generated predictions of that stacked distribution.

As far as I know, the pp_average function does not allow users to generate LOO-CV predictions. But I wondered if obtaining these predictions is as easy as running loo_predict on all my models and weighting the nsamples argument by outputs from model_weights (see this topic).

Thanks for any assistance!

  • Operating System: Windows 11
  • brms Version: 2.16.3