statsmodels.tsa.stattools.pacf_yw#

statsmodels.tsa.stattools.pacf_yw(x, nlags=None, method='adjusted')[source]#

Partial autocorrelation estimated with Yule-Walker

Parameters:

xarray_like: The observations of time series for which pacf is calculated.
nlagsint, optional: Number of lags to return autocorrelation for. If not provided, uses min(10 * np.log10(nobs), nobs - 1).
method{“adjusted”, “mle”}, optional: The method for the autocovariance calculations in yule walker.

Returns:

ndarray: The partial autocorrelations, maxlag+1 elements.

See also

statsmodels.tsa.stattools.pacf: Partial autocorrelation estimation.
statsmodels.tsa.stattools.pacf_ols: Partial autocorrelation estimation using OLS.
statsmodels.tsa.stattools.pacf_burg: Partial autocorrelation estimation using Burg’s method.

Notes

The value at lag k is the last coefficient of the AR(k) model estimated by yule_walker. All lags are computed from one autocovariance sequence using the Levinson-Durbin recursion. yule_walker is used for the lags where the recursion is not known to be stable, which are those where the Toeplitz matrix of the autocovariances is not positive definite.