I have a set of time series which exhibits no autocorrelation but the variance is not constant.

enter image description here

I remember that one of the requirements of cross-correlation to be meaningful is weak stationarity (we use a constant mean and variance in its computation).

In my case, I'm trying to find similar groups of time series, if I use cross-correlation to measure their similarity, two time series will be similar if they follow similar geometric profiles over time, independently of their magnitude. I think that alone makes sense without requiring stationarity.

Can I use it for the time series above and obtain meaningful results in terms of similar geometric profiles? I think stationarity would be only required if I wanted to use the cross-correlation for other inferential purposes. Is this reasoning correct?


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.