I have following question:

When running the ccf command in R, I retrieve results for the cross-correlation of two time series. The data is monthly data, 54 observations, i.e. 5 years. Though the time series are stationary, I assume it's wrong to make a "normal" corr() analysis.

The Autocorrelation results of ccf (ACF) yield that there's a significant correlation at lag 3. That also somehow resonates with results I later find in an impulse response.

However, when applying VARSelect, I wonder why the proposed lag is ranging from 1 to 11 across variables. Shouldn't ACF somehow correspond with information criteria (AIC, BIC), or is the proposed optimum lag length totally independent from autocorrelation?

  • $\begingroup$ This is an interesting question, but you may want to rephrase it somewhat so that it isn't tied as strongly to R terms / code for people who aren't familiar w/ R but may know the answer, or have a similar question in the future. $\endgroup$ – gung Oct 27 '16 at 18:22

Your Answer

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

Browse other questions tagged or ask your own question.