Spider ChartsHi!
I've scoured the community and found a couple of old scripts and add-ins on the topic (or related topics) of spider plots. I was wondering if anyone had access to the code mentioned in the kraft customer success story: https://www.jmp.com/en_us/customer-stories/kraft.html. Was that something that they developed without help from our end?
Thoughts?
M
Re: Spider ChartsI'll agree graphs need to be read and not just looked at, but I'd also say it's not quite like reading words in one important way. Reading graphs relies on eons of evolution prewired in our vision systems. A lot of the vis perception research is in understanding the prewiring so we can make better use of it. Along the way, many perceptual illusions were discovered which highlight how difficult it is for the conscious brain to overrule the visual system. My eyes still get fooled by line-length illusions, for instance.
There needs to be a big pay-off for going against the hard-wiring.
Re: Spider ChartsHey, mark! That’s one of the ones I found in the community. I’m working on taking that code and turning it into something that will overlay plots on top of each other.
Best,
M
Re: Spider ChartsWe forget that we all had to learn how to read plots and graphics just like we had to read words. You learn how to read these plots if you need them. Granted they are not as easy to read or interpret but they work well when you practice.
Re: Spider ChartsHow about this solution?
(Sorry Xan, but you better things to do with your time than code something to shoot down by yourself!)
Re: Spider Charts@Jeff_Perkinson: Yeah - I'm not a fan of them, either. But, in the sensory analysis vertical, they are what people are used to looking at. They use them more for @XanGregg's second example (with multiple variables). I'm willing to put in the effort in the short term to get them onto JMP - and then I can start showing them how awful they are :smiling_face_with_horns:. Ultimately, I need to get them onto some of the multivariate visualizations, but this is the first step.Re: Spider ChartsMike warned me at Discovery that he was having to field a customer request for a spider plot, even though it's generally a bad visual for analytics. They're not terrible (like a truncated bar chart) -- they're just tricky to read (like a 3D bar chart).
I wrote some about them in a blog post https://community.jmp.com/t5/JMP-Blog/One-less-radar-chart-or-be-square-for-Pi-Day/ba-p/37089 which includes a link to a paper that tested their utility in a user experiment. My example was for a continuous cyclic variable (a true angle), which is argubly the best case for angular plots.
A more common use is for a categorical angle (each angle is a separate variable). In those cases, reading the visual becomes more of a gestalt pattern recognition task: is the overall shape round or spiky or off-center or ...? It may work in some cases with careful variable ordering, but a cartesian graph will work better in general.
Btw, the first comment to my blog post was "how do I make a radar/spider in JMP?"! People like them.
Another comment mentions an add-in Clay made, which may be still useful. https://community.jmp.com/t5/JMP-Scripts/Star-Plot-Script/ta-p/21371#U21371
Re: Spider ChartsI'm curious to hear @XanGregg's thoughts about spider charts and which graph types JMP supports that he thinks are a better visualization and why.
Re: Spider ChartsThink you are out of luck. To the best of my knowledge, all of the develoopment for Kraft was done by David Rose in the UK (then a SAS Consultant). He has since moved on, but Robert and/or Phil may have had some more recent contact. He's been to Discovery Europe, for instance.
