models.Elpd

An ELPD estimate from elpd_loo or elpd_waic.

Usage

models.Elpd()

Attributes

Name	Description
elpd	Expected log pointwise predictive density, summed.
ic	The information criterion, -2 elpd (LOOIC or WAIC).
k_threshold	PSIS-LOO only: min(1 - 1/log10(S), 0.7); observations with a
p	Effective number of parameters, lppd - elpd.
pareto_k	PSIS-LOO only: the fitted Pareto shape per observation.
pointwise	ELPD per observation.
se	Its standard error, sqrt(N var(pointwise)).

---

elpd

Expected log pointwise predictive density, summed.

elpd: float

---

ic

The information criterion, -2 elpd (LOOIC or WAIC).

ic: float

---

k_threshold

PSIS-LOO only: min(1 - 1/log10(S), 0.7); observations with a

k_threshold: float | None

larger pareto_k are unreliable.

---

p

Effective number of parameters, lppd - elpd.

p: float

---

pareto_k

PSIS-LOO only: the fitted Pareto shape per observation.

pareto_k: list[float] | None

---

pointwise

ELPD per observation.

pointwise: list[float]

---

se

Its standard error, sqrt(N var(pointwise)).

se: float