Predictive model comparison

Compare multiple models using predictive accuracy estimated using PSIS-LOO-CV. The mc argument is generated using PosteriorStats.compare.

using ArviZPythonPlots, ArviZExampleData, PosteriorStats

use_style("arviz-variat")

models = (
    centered=load_example_data("centered_eight"),
    non_centered=load_example_data("non_centered_eight"),
)
mc = compare(models)
pc = plot_compare(mc)

See plot_compare.