plotting 2 or more timeseries datasets in an x/y plot for correlation

[David Luu]

Similar to these non-D3 posts:

How to plot different timeseries data against each other?

https://www.mathworks.com/matlabcentral/newsreader/view_thread/294460?requestedDomain=www.mathworks.com

https://www.mathworks.com/matlabcentral/answers/109839-how-do-i-plot-one-timeseries-vs-another-from-the-command-line

Say I want timeseries X as my x-axis in the x/y plot (with the x values being the y's in t(X)), and all the timeseries Yn as the Y lines in the x/y plot, where the t in t(X) = t in t(Yn). And/or I may have a summation case, where instead of plotting individual Yn lines, I sum their values to plot a single Y for X.

How is this best done with d3.js? I'm working on it right now, and wanted to get some advice on it while I'm figuring things out.

[Paul Hobson]

I think you're going to need to pre-process your data outside of D3. 

Faced with this problem, I would use pandas to resample each timeseries to common and slightly coarser timesteps: http://pandas.pydata.org/pandas-docs/stable/timeseries.html#resampling

Another approach I would take is to make a Q-Q or P-P plot to compare the distributions of the data.

http://www.itl.nist.gov/div898/handbook/eda/section3/eda33o.htm

The Q-Q plot won't require resampling if both of your samples have the same length.

See this discussion for more info: https://github.com/statsmodels/statsmodels/issues/2896#issuecomment-238033335

-paul

--
You received this message because you are subscribed to the Google Groups "d3-js" group.
To unsubscribe from this group and stop receiving emails from it, send an email to d3-js+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.