VARMA._innovation_block#

VARMA._innovation_block(innovations, start)[source]#

Stack u_{t-1}, ..., u_{t-q} for rows start onward.

Parameters:

  • innovations (NDArray[double]) – An (nobs, k) innovation series aligned with the sample.

  • start (int) – First time index to build a row for; must be at least q.

Return type:

NDArray[double]

Returns:

An (nobs - start, k * q) block, lag-major.

Example

>>> import numpy as np
>>> model = VARMA(np.random.default_rng(0).standard_normal((200, 2)), order=(1, 2))
>>> innovations = np.arange(400.0).reshape(200, 2)
>>> block = model._innovation_block(innovations, 10)
>>> block.shape, block[0].tolist()
((190, 4), [18.0, 19.0, 16.0, 17.0])