CODEWITHSUNDEEP
python-3.xtime-seriesstatsmodels
Perl Del Rey
Perl Del Rey

Reputation: 1059

adfuller number of lags returned

I have the following code snippet for performing the adfuller test on my time series

def dickey_fuller(series):
    X = series
    result = adfuller(X, regression='ct')
    print('ADF Statistic: %f' % result[0])
    print('p-value: %f' % result[1])
    print('Lags used: %d' % result[2])
    print('Critical Values:')
    for key, value in result[4].items():
        print('%s: %.3f' % (key, value))
    print('-----------------------------------------------')

Those are the adfuller values by default:

statsmodels.tsa.stattools.adfuller(x, maxlag=None, regression='c', autolag='AIC', store=False, regresults=False)

As cited in the statsmodels docs, when we have autolag='AIC' it will choose the number of lags that minimize the information criterion.

The following is a sample output of my run of this test on my data:

ADF Statistic: -7.359845
p-value: 0.000000
Lags used: 7
Critical Values:
1%: -4.021
5%: -3.441
10%: -3.145

I am interested in including the appropriate number of lags as column in my data, and here the used number of lags is 7. In this case, shall I take all lags from 1 up to 7 or shall I take only the 7th lag?

Or in other words, is the returned number of lags the lag length or the best lag taken ?

Upvotes: 0

Views: 1289

Answers (1)

Nathan Furnal
Nathan Furnal

Reputation: 2410

Per the docs it returns the number of lagS used. But typically, when talking about such models, somethink like : AR(2) refers to an autogressive model using two lags. If you use some lags like every other seventh or twelveth lag, it's a seasonal component.

Upvotes: 1

Related Questions