VARMA._burn_for

VARMA._burn_for(order)

Leading observations lost across all three passes.

The second pass needs the long autoregression’s residuals plus q lags of them, so the burn is long_order + q whenever that exceeds the autoregressive order; it is what the base class’s minimum-sample check reads.

Parameters:

order (int) – Autoregressive order under consideration.

Return type:

int

Returns:

max(order, long_order + q).

Example

>>> import numpy as np
>>> rng = np.random.default_rng(0)
>>> VARMA(rng.standard_normal((400, 2)), order=(1, 1))._burn_for(1)
21