News for hackers. Stuff that matters.
Re:Downsampling Time Series for Visual Representation [pdf] ... by dabreegster on Tuesday March 09, 2021 @05:09PM (#26405043)
To be more specific, I'm measuring throughput along road segments in a deterministic traffic simulation. Differences would occur when the user modifies the road network. The modifications often have no effect on most roads, but some wind up with more or less traffic than usual at a certain time.

I'm interested in showing differences like "during morning rush hour, there was more traffic here, but it was about the same the rest of the day", so I'm not sure a single correlation as a metric would be best. Ideally I'd plot two line plots and call out differences over time.

Re:Downsampling Time Series for Visual Representation [pdf] ... by 1996 on Tuesday March 09, 2021 @05:19PM (#26405159)
There is going to be complex behavior (adapting to traffic by routing around it) that a simulation will poorly predict

Also, you will have correlation between adjacent segments of the network- and the idea of extracting "rush hours" brings its own set of problems.

Consult with a statistician familiar with geographical models. I'm sorry I can't help much more than that. It's very different from finance.

Re:Downsampling Time Series for Visual Representation [pdf] ... by dabreegster on Tuesday March 09, 2021 @05:25PM (#26405232)
Surprises from emergent behavior are some of the most satisfying things about working on this.

I'm not trying to extract rush hour patterns or anything quantitatively, just show two timeseries plots to the user and have them figure out what's going on.

Thanks for the ideas! Lots of new things to research. Stats isn't my background at all.