Given a Bayesian "projpred" or "brms" model, computes linear and full
posterior predicitons.
Usage
compute_predictions(model, seed = NULL, ...)
Arguments
- model
The model
- seed
Optional seed for reproducibility
- ...
Unused. There to allow for compatibility with run_scoring_rule_pipeline.
Value
Tibble with two columns:
- epred
Linear prediction on the logit scale
- ppred
Posterior prediction on the outcome scale