Closed light-and-salt closed 8 years ago
Then you are wasting a lot of bin for empty values? On Thu, Sep 8, 2016 at 7:49 PM Zening notifications@github.com wrote:
[image: image] https://cloud.githubusercontent.com/assets/822034/18373883/db6c621e-75fc-11e6-8f0f-3d05a5966fbc.png
If the specified view Horsepower includes zero, then the binned Horsepower in the related view should also start from zero.
Seems like a good cross-view comparison scenario for my consistency project :)
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/uwdata/voyager2/issues/260, or mute the thread https://github.com/notifications/unsubscribe-auth/AAGypforMurEy2_AiFmPL2ZIPe408HVaks5qoMk5gaJpZM4J4pjC .
In this case if the bin plot start from zero, it will probably waste two bins. But that's just like the specified view, which starts from zero and has some whitespace on the left next to the y axis. When I saw this screen I think the related views are supposed to be compared (or related) to the specified view. If the bin axis start from zero, the green, blue and orange points will be in the same relative position as the specified view, so I think it will be easier to relate the bin plot to the specified view.
If you data range from 1000-3000, you're wasting 67% of the bins as well. At that point, I'm not sure how much consistency I'd want to preserve. But yeah, it's a design decision question.
I'd say it's probably an overkill to solve this problem as people can always read the label.
If you data range from 1000-3000, you're wasting 67% of the bins as well.
If your data range from 1000-3000, then the specified view shouldn't include zero in the first place, and all the related views shouldn't include zero either. So the specified view and the related views can still be consistent without sacrificing too much space efficiency.
then the specified view shouldn't include zero in the first place
That's not true. Including zero is important for determining ratio in many cases.
OK, if supporting ratio comparison is one of the (most important) tasks of the original (unbinned) plot, and the primary task of the bin plot is to show distribution, then I think it's a totally valid rationale to overwrite consistency.
Each chart should do its own job as well as related to other charts in the same set.
If the specified view Horsepower includes zero, then the binned Horsepower in the related view should also start from zero.
Seems like a good cross-view comparison scenario for my consistency project :)