arviz_stats.base.array_stats.loo_mixture

array_stats.loo_mixture(ary, obs_axes, chain_axis=-2, draw_axis=-1, log_jacobian=None)

Compute mixture importance sampling LOO (Mix-IS-LOO).

Parameters:

aryarray_like

Full log-likelihood array.

obs_axestuple of int

Axes corresponding to observation dimensions.

chain_axisint, default -2

Axis for chains.

draw_axisint, default -1

Axis for draws.

log_jacobianarray_like, optional

Log-Jacobian adjustment for variable transformations.

Returns:

elpd_iarray_like

Pointwise expected log predictive density.

p_loo_iarray_like

Pointwise effective number of parameters.

mix_log_weightsarray_like

Mixture log weights.