I have a time series question that is probably answered by these posts (1, 2) but perhaps I'm not thinking about it in the right way.

I have 3 time series: $A$, $B$, and $C$. The three time series measure related quantities but have different units/scales. I have good reason to believe that both $A$ and $B$ are essentially controlled by $C$. $C$ Is quite periodic, and and $A$ and $B$ look pretty periodic but noisier.

Is there a statistical metric for comparing how well the periodicity of two time series match? The posts on cross-correlation and Pearson coefficient suggest that the ARIMA structure of each time series should be discerned independently and the 'within-series' structure should be removed so that the error terms of each time series can tested for correlation while avoiding 'spurious correlation'. However, I think that might be the opposite of what I want to do---I know the cause of periodicity in $C$, and I want to see how well the observed periodicity in $A$ and $B$ match that observed in $C$.

Does it still make sense to remove the 'within-series' structure in this case? Is correlation the right way to be thinking about this problem, or should I be thinking about e.g. Fourier transforms to express periodicity and compare transformed metrics (something I have never done before)?


1 Answer 1


I ended up using the Lomb-Scargle periodogram (R package lomb) to do this. The Lomb-Scargle periodgram can be computed for irregular series. Also, the package can compute p-values for the most significant peaks. I had a (known and not interesting) linear trend in each series that I removed prior to computing the periodogram for each series, but otherwise kept the periodic structure of the series and didn't remove any other "within-series" structure (since that was what I was interested in!).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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