VARMA._long_order

VARMA._long_order()

Order of the stage-one autoregression.

Grows like the square root of the sample, which dominates any fixed (p, q) asymptotically and keeps the residuals consistent for the innovations, capped at what the sample can identify so that the first stage never fails on a design its own target cannot support.

Return type:

int

Returns:

max(p + q, ceil(sqrt(nobs))), capped by the largest order the sample’s rows and width allow, and at least one.

Example

>>> import numpy as np
>>> rng = np.random.default_rng(0)
>>> VARMA(rng.standard_normal((400, 2)), order=(1, 1))._long_order()
20
>>> VARMA(rng.standard_normal((30, 2)), order=(1, 1))._long_order()
6