arviz_stats.base.array_stats.loo_score

array_stats.loo_score(ary, y_obs, log_weights, kind='crps', chain_axis=-2, draw_axis=-1)

Compute CRPS or SCRPS with PSIS-LOO-CV weights.

Parameters:

aryarray_like

Posterior predictive samples.

y_obsarray_like

Observed values.

log_weightsarray_like

Pre-computed PSIS log weights.

kindstr, default “crps”

Score type, either “crps” or “scrps”.

chain_axisint, default -2

Axis for chains.

draw_axisint, default -1

Axis for draws.

Returns:

scoresarray_like

Score values (negative orientation for maximization).