how can I detect unit root/non-stationarity?
let’s say that there is one dependent variable and two independent.
if we can reject the two independent variable, does this mean that unit root exists for the dependent variable?
additionally, what is the impact on the result?
thanks in advance
To detect a unit root is the dicky-fuller test.
After you first difference the data test b1 - 1 for significance.
If it’s statistically significantly different from zero you have a unit root and therefore no mean reverting level.
If it’s not different than zero then you’re okay and the AR model has a mean reverting level which are two requirements for an AR model to be cov. Stationary: 1) No unit root 2) has a mean reverting level.
This is my no cheating no flashcard looking answer to the question which could be horribly wrong though…
To detect Unit Root you have to use the Dickey-Fuller test (Dickey-Fuller Engle-Granger is for cointegration between 2 time series)
Your test static would be b1 - 1 or sometimes referred as g
H0 = Unit Root / Ha = Cov.Stationary
So if g (b1 - 1) is statistically significantly different from zero (Reject H0), then b1 would be different form 1 and your time series would be covariance Stationary
If g is not different than zero (don’t Reject H0), then b1 would be equal to 1 and the time series would hve a unit root
Sorry I had it backwards.
Get that out of your system now.
ie a unit root with cointegration is acceptable 
don’t complify that simplicated side right?