I would like to assess the synchrony between two time-series (i.e. do they have the same evolution over time?). I am using Python.
Examples of the time series are in the files attached. The first plot shows time-series that visually do not have similar evolution whereas the second plot shows series that have quite similar evolution (synchrony).
I have considered Dynamic Time Warping to assess the similarity between the time series, but the problem is that we obtain a distance between the time series, which is difficult to interpret. What I would like is rather a standardized number (e.g. between -1 and 1 like a correlation) that would show if the series are synchronous (i.e. they increase or decrease at the same time). This way I could compare the evolution of several time-series more easily even if their global distance (i.e. average value) is different like shown on the first vs. second plot. Is it possible to have such number with Dynamic Time Warping? Or would another method be more appropriate?
I would start by looking a the cross correlation. It also depends on what you want to do with these series. Best, D. Booth
David Eugene Booth Tank you for your response. My objective is to create an algorithm to automatically identify time series that have similar or dissimilar evolution over time (I will have a large number of plots like this to analyze regularly).
Hi Pauline Delhez
Some more information:
https://www.researchgate.net/post/Multiple_cross-correlation_function12
And good luck if your aim is cross-correlation between biomarkers and feeding.
Roberto Molteni I agree but i think it goes deeper. one time series is difficult to identify so to find the the same pattern in multiple series seems much worse. But i found that work has been done on this problem, see : https://www.google.com/search?client=firefox-b-1-d&q=Grouping+simlalar+time+series
Best, David Booth
I agree prof. David Eugene Booth ,
but Pauline Delhez said "I will have a large number of plots like this to analyze regularly".
For that, I also wished "good luck" for her job.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series