arviz_stats.base.array_stats.loo_r2

array_stats.loo_r2(y_obs, ypred_loo, n_simulations=4000, circular=False, random_state=42)

Compute LOO-adjusted R-squared using Dirichlet-weighted bootstrap.

Parameters:

y_obsarray_like

Observed values.

ypred_looarray_like

LOO predictions (same shape as y_obs).

n_simulationsint, default 4000

Number of Dirichlet-weighted bootstrap samples.

circularbool, default False

Whether the variable is circular (angles in radians).

random_stateint, default 42

Random seed for reproducibility.

Returns:

loo_r_squaredarray_like

R-squared samples with shape (n_simulations,).