statsmodels.regression.linear_model.GLSAR.whiten

GLSAR.whiten(x)

For an AR(p) process, the errors are modeled as

\[u_t = \rho_1 u_{t-1} + \cdots + \rho_p u_{t-p} + \epsilon_t,\]

where \(\epsilon_t\) is white noise. The corresponding whitening transformation is

\[\epsilon_t = u_t - \sum_{i=1}^{p} \rho_i u_{t-i}.\]

Whitening using this method drops the initial p observations.

Parameters:

xarray_like

The data to be whitened.

Returns:

ndarray

The whitened data.