I have a time series data for 18 months. To check for stationary I conducted adf test, to which my p value is 0.8. And kpss test has a p value of 0.1 , so at 95% confidence level I fail to reject null in both cases. That means my adf test confirm that data is non stationary whereas kpss test confirms my data as stationary. Why is that happening?


1 Answer 1


For the data you have, neither test is powerful enough to reject the null hypothesis. But it looks like KPSS test is not too far from rejecting it; a little more power (a little more data), and that might have happened. If so, the results of the ADF and KPSS tests would not be contradictory anymore. So I do not see a big disagreement in the test results in your application.

  • $\begingroup$ Thanks Richard. $\endgroup$
    – Manisha
    Dec 4, 2019 at 14:27
  • $\begingroup$ Do we have any other powerful test to have a confirmation? $\endgroup$
    – Manisha
    Dec 4, 2019 at 14:28
  • $\begingroup$ There are some other tests like the ADF-GLS test which is likely more powerful than the vanilla ADF test, but sometimes the sample is simply too small to allow for inference at high levels of confidence. $\endgroup$ Dec 4, 2019 at 16:56
  • $\begingroup$ Okay,thanks so much Richard for your time $\endgroup$
    – Manisha
    Dec 4, 2019 at 18:47
  • $\begingroup$ @Manisha, thank you, I am glad to have helped. $\endgroup$ Dec 4, 2019 at 21:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.