The authors' paper about the method (Hyndman and Khandakar 2008) states that an extended an extended Canova-Hansen test is used, although the helpfile updates this to say than an OCSB test is used.

I have two questions, therefore:

  1. What is an OCSB test?
  2. Is the method of determining seasonal differencing the same for regression with ARIMA errors as it is for ARIMA modelling with no covariates?
  1. OCSB test: Osborn DR, Chui APL, Smith J, and Birchenhall CR (1988) "Seasonality and the order of integration for consumption", Oxford Bulletin of Economics and Statistics 50(4):361-377.

    The change in the algorithm is explained at http://robjhyndman.com/researchtips/forecast3/

  2. When regressors are used, auto.arima() fits the linear model using lm() and applies the unit root tests to the residuals.

  • 2
    $\begingroup$ Thanks very much Rob for your timely answer (on a public holiday in Aus, no less)! $\endgroup$ – fmark Jun 11 '12 at 5:13
  • $\begingroup$ @RobHyndman Is there an implementation of HEGY test for testing of seasonal unit root in R ? or is OSBN preferred ? There is a uroot packgage but as its not on CRAN, dont know how reliable that is. Thanks. $\endgroup$ – Anusha Sep 22 '14 at 7:27
  • $\begingroup$ Please ask a new question, rather than bury questions in the comments to answers. $\endgroup$ – Rob Hyndman Sep 22 '14 at 13:01
  • $\begingroup$ Prof. Hyndman, are the regressors automatically subject to the same degrees of differencing as the ARIMA model (d,D)? $\endgroup$ – Ricardo Coqueiro Sep 5 '15 at 11:15

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.