Econometricians often talk about a time series being integrated with order k, I(k). k being the minimum number of differences required to obtain a stationary time series.

What methods or statistical tests can be used to determine, given a level of confidence, the order of integration of a time series?


There are a number of statistical tests (known as "unit root tests") for dealing with this problem. The most popular is probably the "Augmented Dickey-Fuller" (ADF) test, although the Phillips-Perron (PP) test and the KPSS test are also widely used.

Both the ADF and PP tests are based on a null hypothesis of a unit root (i.e., an I(1) series). The KPSS test is based on a null hypothesis of stationarity (i.e., an I(0) series). Consequently, the KPSS test can give quite different results from the ADF or PP tests.

  • $\begingroup$ Is there a convenient way to determine the order of integration if beyond I(0) or I(1), e.g. I(2) etc.? $\endgroup$ – user45065 Feb 13 '17 at 13:24
  • 1
    $\begingroup$ Difference the data and apply the tests again. $\endgroup$ – Rob Hyndman Feb 13 '17 at 21:25

Also, for some elaborate discussion (including bashing of ADF / PP / KPSS :) you might want to have a look at the book by Maddala and Kim:


Quite extensive and not very easy to read sometimes, but a useful reference.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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