Cointegration test (Econometrics)

[cynthiiiiiaaaaa]

Hi all,
I'm designing a quant model, and wondering whether it is possible to contain two cointegration tests? The design of the quant trading model is:

Test the stationarity through Augmented Dickey Fuller test,

1) just leave a time series (price data of a stock) if it is I(0) process, and because it is stationary, then test this with Augmented Dickey Fuller Cointegration test, or

2) If a time series is I(1) process, which means non-stationary, then put it through the Engle Granger test (which tests two combination of non-stationary time series)

It would be much appreciated if anyone can leave a comment on this thread.
if any of my logic is incorrect or wrong, please let me know.

Thanks in advance

 

[bcdp]

Hi cynthia,
I've recently studied the literature about cointegration and your approach is exactly the one adopted by Clive Granger itself.
He first tests if the process is I(1) and then run the cointegration tests on several processes, in order to detect hidden correlation.
He finishes the analysis by developing an error correction model which takes care of (previously detected) correlations.

References:

1. Unit Root test - "Treasury Bill Yield Curves and Cointegration", Anderson, Granger and Hall (1990)
2. Cointegration and ECM - "A Cointegration Analysis of Treasury Bill Yields", Hall, Anderson, Granger (1992) (see attachement)

If you need it I can also provide you the .pdf "Essays in Econometrics: Collected Papers of Clive W. J. Granger"