Adf interpretation

[Toni2]

below is my ADF test for regression residuals. I use alpha=5%

My interpretation is : my residuals are non-stationary as my model does not take into account the 4 lags suggested by ADF. In other words, my residuals would be stationary only if i take in consideration the 4 lags in the model

Is this correct?

Lag 4 , ADF = -2.922, p-value = 0.0428

/* Perform the ADF test on the residuals with 4 lags */
proc autoreg data=residuals;
   model resid / adf=(noint) nlag=4;
run;

[SASCom1]

Hi @Toni2 

I am afraid I do not follow the logic in your statement: "my residuals are non-stationary as my model does not take into account the 4 lags suggested by ADF. In other words, my residuals would be stationary only if i take in consideration the 4 lags in the model.'

When you perform ADF stationarity test on a series, you need to make the decision on the number of lags to be included in the test. Once you have made the decision on the number of lags, you can request the ADF test with the requested number of lags using ADF = (number ) option. For example, if you decide you should include 4 lags in the ADF test, then you specify stationarity = (adf=(4)) option in the MODEL statement. If you do not specify ADF = ( number), then by default, PROC AUTOREG perform the test using 0 lag.

The null hypothesis for the ADF test is unit root, so you reject the null hypothesis with small p values. You look at the row corresponding to the model you choose, and the number of lags you decided, and reject the null hypothesis for small p values, and do not reject the null of unit root for large p values.

The following paper by Dr. David Dickey provides helpful information on various issues regarding stationarity in time series which may be helpful to you:

https://support.sas.com/resources/papers/proceedings/proceedings/sugi30/192-30.pdf

I hope this helps.