Module criteria 

Classical information criteria over nested AR fits.

Innovation variances come from the Yule–Walker/Levinson recursion (same convention as R’s ar.yw), so every order is costed in one O(K²) pass. An OLS-conditional variant is planned so that the comparison with the OLS-based Fine–Hwang statistic is apples-to-apples.

Structs

IcResult

Result of ic_order.

Enums

Criterion

Penalised-likelihood criteria (all “smaller is better”).

IcMethod

How the innovation variances of the nested AR fits are estimated.

Functions

default_max_order

Default maximum order used by R’s ar(): min(n-1, 10 log10 n).

ic_order

Select an AR order by minimising criterion over 0..=max_order, with Yule–Walker innovation variances. See ic_order_with for other methods.

ic_order_all

All criteria in one pass (shares the Levinson recursion).

ic_order_with

Select an AR order by minimising criterion over 0..=max_order, with the innovation variances of the nested fits estimated by method.