I noticed that the crosscorrelation plot shows both positive and negative lags. I have a situation where only lag -1 has a significant crosscorrelation, but I don't think that is actionable?? I've prewhitened and taken the first difference of the response series and the explanatory variable. Thanks!
The cross-correlation test of two time-series data sets involves many calculations of the correlation r by time-shifting the one data set relative to the other data set. Each shift is called a "lag", and the lag time is simply the sampling period of the two time-series data sets. A typical cross-correlation graph shows enough lags in both negative and positive directions to show the cyclical relationship of the two sets of data.
Does that help?
Thank you, that does help shed light on the purpose of the crosscorrelation graph. In lieu of any cyclical relationship, is the significant crosscorrelation at lag -1 actionable? I am interpreting to results to mean that the explanatory variable is not useful.
Not sure of your specific situation, but there are cases where a crosscorrelation value of -1 is of practical usage. See for example: http://jds.fass.org/cgi/reprint/75/7/1891.pdf. Taken from there: "The significant negative crosscorrelation between milk yield and cholesterol on the same day suggests that serum cholesterol may be removed in milk or, more probably, that alterations in metabolism required for increased milk yield result in lower blood cholesterol."
Message was edited by: udo@sas