Dicky Fuller Test and autocorrelation in the Residuals

[Henning Meyer]

Hi guys,

I currently have a problem in interpreting the output of the Dickey Fuller test when working on a time series.
How can I figure out if the "dfuller" command accounts for possible autocorrelations in the residuals?
Do I have to test them separately using a Breusch-Godfrey test?

Thanks so much for your help!
Henning

[John Riveros]

Henning, Dicky Fuller has the null hypothesis of unit roots. if the test of your p-value equals to 0.0000 means you'll reject this H0. and accept by definition the alternative hypothesis which is variable is Stationary (which is what is desirable).

However, i would perform some graphs over the residuals. for example. scatter the residual against it's lagged residual. (if the residuals seems to have a linear correlation then autocorrelation of first order it's happening). You can perform as well some LM (Breush-Godfrey) test over the residuals of the which is why the estat bgodfrey helps. in order to check it. So if p-value it's equal to 0.000 you'll reject the null hypothesis of no autocorrelation at a respective lag. and accept the autocorrelation.

[Joro Kolev]

Further, "dfuller performs the augmented Dickey-Fuller test that a variable follows a unit-root process. " The word "augmented" comes from the fact that you are supposed to include enough lags of the differenced variable, until there is no autocorrelation in the residuals.

So the answer to your question is No, dfuller does not account for possibly autocorrelation in the residuals. You are supposed to say explicitly to dfuller how many lags of the differenced variable you want to have included, so that there is no autocorrelation in the residuals.