arch.covariance.var.PreWhitenedRecolored.resid¶

property PreWhitenedRecolored.resid : ndarray[tuple[int, ...], dtype[float64]] | DataFrame¶

The residuals of the VAR used to prewhiten x.

Returns:¶

The nobs - q by nvar residuals, where q is the second element of order, so that the first q observations of x are lost. A DataFrame with the dates or index of the final observations of x and the same columns is returned if x is a DataFrame or a Series, and an ndarray otherwise.

Return type:¶

numpy.ndarray or pandas.DataFrame

Notes

The kernel estimator is applied to these residuals and not to x, with the autocovariances computed without demeaning them. They can be used to check that the prewhitening was adequate. If the residuals are still serially correlated, a larger lags or max_lag should be used. When the order is 0 and center is True the residuals are x less its mean, and when the order is 0 and center is False they are x. If center is True the VAR includes a constant, so the residuals have mean zero. The residuals do not require the VAR to be covariance stationary.