Open marcdotson opened 4 years ago
Moved draft from hmnl-comparison
in quant-seminar
.
Possible use of posterior along with tidybayes and bayesplot.
And what about tidymodels integration?
Include some investigation around LOO:
Rather than leave-one-out cross-validation, k-fold cross-validation (leave-on-group-out) would be a good fit for hierarchical models. See Merkel, Furr and Rabe-Hesketh (2018) for an approach using quadrature integration.
Interesting discussion on recovering parameters.
Jim's take on specific prior predictive checks for choice models.
What Bayes options are there in tidymodels?
See the current draft of this idea, as well as the latest principled Bayesian workflow by Michael Betancourt.
How might this relate to the use of Stan as an engine in tidymodels?