models.lppd()

In-sample log pointwise predictive density,

Usage

models.lppd(log_lik)

sum_i log(mean_s p(y_i | theta_s)).

Parameters

log_lik: list of list of float

One row per posterior draw, one column per observation.

Returns

float

Examples

>>> import math
>>> from prospicio.models import lppd
>>> round(lppd([[math.log(0.5)], [math.log(0.25)]]), 12) == round(math.log(0.375), 12)

True