topic Alternative to Box Plots for large datasets? in App Development

Alternative to Box Plots for large datasets?

dwqlik82 — Wed, 29 Mar 2023 14:21:29 GMT

Hi,

Was just wondering if people had a good alternative that would generate something similar to a box plot but would work for large-ish data sets (from my testing the dreaded *showing a limited dataset appears about 31000 rows ish depending).

I've built a stacked bar combo chart using something like the default Tukey rules (used combo so i could create a transparent offset) but it seems a bit clunky (and hides outliers):

currently have both within a container with conditional hide/show depending on the rows in the current selection, was just wondering if other people had hit similar issues and other ways people had handled it (also if anyone has a decent idea on the actual row limits for the limited data set appearing that would be great)

thanks

Dale

Re: Alternative to Box Plots for large datasets?

marcus_sommer — Thu, 30 Mar 2023 07:29:40 GMT

Alternatives may be not to use a single object else several ones which are placed side by side and/or overlapping each other. Adjusting all needed layout-features and trying to synchronize multiple objects to each other is not always easy and some disadvantages may remain - in regard to show/hide, exports/prints and/or using any popup-features and/or direct selecting within the objects ... Also thinkable is to use a straight-table with mini-charts and/or colored numbers of ascii chars ... or maybe using line-charts ...

More helpful may be to re-consider the entire approach regarding to the most important information which should be provided with the chart and if there may other and more suitable views.

Re: Alternative to Box Plots for large datasets?

dwqlik82 — Thu, 30 Mar 2023 14:12:11 GMT

Cheers, i'd guessed as much but hoped I'd missed something, I found the box and whisker chart good for comparing median and spread of across multiple dimensions and presenting it in a relatively easy to understand view (in my case looking at the time between occurrences of X then splitting it either by Department, Date, Weekday etc to try and let people investigate if wider variability is focused in certain areas). I may see if just expanding out my pseudo box plot based on a combo chart and making it a bit more readable may work. Nuclear option is building an extension to do it but that is usually far more effort than its worth 🙂

Thanks again