In order to determine the right time lag for the Granger causality test, I applied a cross correlation function (CCF) among EU and National legislation and the CCF figure is attached. I would be grateful if you could kindly help me in resolving the following concerns:

  • I am wondering what time-lag you recommend based on the attached figure?
  • To find the greatest value in CCF figure, should I consider absolute values or just the raw values?
  • If we consider the negative values in CCF Figure, then I assume the result of Granger causality test would show the negative causality, correct?
  • I am interested in one side of the causality (Effect of EU on National and Not the vice versa), so, should I only consider the bound [-5...0]?
  • $\begingroup$ You need more reputation to post images, but you are able to post data and/or printed output from the software you used, such as numeric tabulation of CCF. $\endgroup$
    – Nick Cox
    Aug 7, 2013 at 11:00
  • $\begingroup$ The OP says the figure is attached but there is no figure here. It seems to be lost during the editing. $\endgroup$
    – Metrics
    Aug 12, 2013 at 2:06

1 Answer 1


Does the lag identification procedure have to be applied using the cross correlation? If not, I would recommend the function VARselect() from the package vars in the R programming environment. This function suggests different optimal lags according to some information criteria (AIC, HQ, SC and FPE). Once you have chosen the lag, you can perform a Granger causality test using another function from the same package causality().

  • $\begingroup$ What language is this and how do you know the OP is using it? $\endgroup$
    – Nick Cox
    Aug 7, 2013 at 15:59
  • $\begingroup$ The language is, of cource, R. $\endgroup$ Aug 7, 2013 at 16:12
  • 1
    $\begingroup$ As neither R nor any other language is a declared default on this site, it is a matter of courtesy as well as information always to explain the language being used. The second half of my question remains.... $\endgroup$
    – Nick Cox
    Aug 7, 2013 at 16:15
  • $\begingroup$ I can only suggest a language I use... $\endgroup$ Aug 7, 2013 at 16:25
  • $\begingroup$ Naturally I agree with that. However, the question is not about software, but about methods, so in principle a fair answer need not mention software at all. In practice we can't see the OP's graphs, but that's a different point. $\endgroup$
    – Nick Cox
    Aug 7, 2013 at 16:29

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.