statsmodels.tsa.stattools.pacf¶

statsmodels.tsa.stattools.
pacf
(x, nlags=40, method='ywunbiased', alpha=None)[source]¶ Partial autocorrelation estimated
 Parameters
 x1d array
observations of time series for which pacf is calculated
 nlagsint
largest lag for which the pacf is returned
 methodstr
specifies which method for the calculations to use:
‘yw’ or ‘ywunbiased’ : YuleWalker with bias correction in denominator for acovf. Default.
‘ywm’ or ‘ywmle’ : YuleWalker without bias correction
‘ols’ : regression of time series on lags of it and on constant
‘olsinefficient’ : regression of time series on lags using a single common sample to estimate all pacf coefficients
‘olsunbiased’ : regression of time series on lags with a bias adjustment
‘ld’ or ‘ldunbiased’ : LevinsonDurbin recursion with bias correction
‘ldb’ or ‘ldbiased’ : LevinsonDurbin recursion without bias correction
 alphafloat, optional
If a number is given, the confidence intervals for the given level are returned. For instance if alpha=.05, 95 % confidence intervals are returned where the standard deviation is computed according to 1/sqrt(len(x))
 Returns
 pacf1d array
partial autocorrelations, nlags elements, including lag zero
 confintarray, optional
Confidence intervals for the PACF. Returned if confint is not None.
See also
statsmodels.tsa.stattools.acf
,statsmodels.tsa.stattools.pacf_yw
,statsmodels.tsa.stattools.pacf_burg
,statsmodels.tsa.stattools.pacf_ols
Notes
Based on simulation evidence across a range of loworder ARMA models, the best methods based on root MSE are YuleWalker (MLW), LevinsonDurbin (MLE) and Burg, respectively. The estimators with the lowest bias included included these three in addition to OLS and OLSunbiased.
YuleWalker (unbiased) and LevinsonDurbin (unbiased) performed consistently worse than the other options.