4

I am trying to do cointegration analysis between two variables. I first used the standard Dickey-Fuller and Phillips-Perron tests; they concluded my variables were I(1). I then did cointegration and got positive results.

However, I have now gone back to the Dickey-Fuller test and included a trend term. This resulted in rejection of the null of non-stationarity for one of my variables.

Does this mean I can't continue with cointegration analysis?

Richard Hardy
  • 54,375
  • 10
  • 95
  • 219
Axl Lopez
  • 41
  • 2
  • see https://stats.stackexchange.com/questions/44647/which-dickey-fuller-test-for-a-time-series-modelled-with-an-intercept-drift-and –  Dec 11 '17 at 20:44

1 Answers1

-2

Well firstly you should use the Augmented Dickey Fuller.

Then, no you can't continue with cointegration if you have a stationary series.

Look at your series, if one is trend stationary then you can't continue with the cointegration test!

J.Doe
  • 1
  • 1
  • Keele, L., & Boef, S. D. (2006). [Not Just for Cointegration: Error Correction Models with Stationary Data](https://pdfs.semanticscholar.org/32fa/a6da80e2e8361bf6c650e11a2033d8b3b88b.pdf). Retrieved from http://www – Alexis Apr 15 '19 at 03:19