The short answer is that Tableau doesn’t support 3d graphs because they aren’t consistent with best practices for presenting data in a way that it can be accurately understood. I completely agree with that as a general rule; I absolutely hate 3d graphs like the ones you can easily produce in excel. If that is the goal you can export a crosstab as a csv and go that route. In Tableau, a better solution might be to plot several histograms as line graphs on the same axis with something on the color shelf to distinguish between them. That would solve the space issue and allow these to be compared to one another.
Given that disclaimer, 3d is something I have given some thought, not so much as a data scientist but as a mathematician. The truth is that we don’t actually see in 3d, it is constructed in our brain based on a few clues:
- Things that are further away get smaller. This might be called perspective.
- With a few exceptions, we only see the closest thing in any direction, so if one object obscures another we understand something about its relative position (shadows can also give some info, but phrasing this broadly enough to include this makes it overly complex). This type of clue is relational.
- Most of us have 2 eyes, each of which sees things slightly differently. This is known as parallax. This is what makes movies 3d, and is the most interesting to me. Basically each of us has a dedicated processer in our brain for doing trigonometry using the angles at which each eye sees an object and the fixed distance between our eyes (2 angles & 1 side is enough to reconstruct a complete triangle, but good luck catching a baseball if you think about it).
Excel uses 1 & 2 to make 3d-ish graphs, but the experience is the same even if you close one eye (which, like color blind palates, might be worth considering if you’re distributing your viz). Excel style 3d may have added some flash, before anyone knew better, but it doesn’t really add substance. Not worth distorting or obscuring data. Adding #3 probably still isn’t worth it in general, but wouldn’t it be exciting to see a viz jumping out of the screen once in a while (and who knows what might stir the idea that is worth pursuing).
The problem with creating parallax is that it means each eye needs to see the image differently. Special equipment or training (probably both) is required to really pull this off. In most cases this is too expensive and complex for someone with a casual interest. There is one noteworthy exception to this, called Chromadepth.
A few years ago I noticed a box of 3d sidewalk chalk. I was intrigued enough to pick it up and found it pretty easy to create images on my chalkboard with distinct 3d layers. The glasses are basically weak prisms scattering light slightly based on color, so reds & yellows appear closer to the viewer while Blues and Purples appear further away. In my experience, they require a dark background (Yay dark maps!) and work better with just a few distinct colors that are ordered by frequency. Both of these are easy to do in Tableau. I’ve seen these glasses all over lately, even in my local pharmacy (Crayola seems to be the main distributer, though usually only in children’s sizes). Look for things like 3d coloring books with glasses that aren’t red/green. A couple pairs of these and you’ll find some of your vizs may already be 3d. Also, unlike other 3d technologies, the image doesn’t suffer if you don’t have special glasses, you just don’t see the extra dimension.
In case all the pharmacies in Seattle sell out, or you’re confused what to look for, here is a link:
First I can't imagine there is any way I could possibly beat "3D sidewalk chalk"! Noah is the boom. But if you can provide a simple data set you want to chart I have an idea that you might be able to do it in SketchUp. Then you could fly around your histogram [and fool yourself] to your heart content. And best of all no joystick required.
Thanks Shaw, assuming the boom is a complement.
I thought we were looking for Tableau solutions. You could definitely do this in sketchup. But if you're going that far, and throwing best practices to the wind, you may as well download your favorite baseball stadium from their 3d warehouse and put your histograms in the outfield. Then you fly in and spin the whole thing around while you discuss the significance and then fly on to the grand canyon for your next bullet point.
Let me know if you pull off some 3d graphs, I'd like to see them. A word of caution, though, like passing out 3d glasses, I expect this would end with an audience that has no idea what you actually talked about, unless it is making 3d graphs, and think you're a showoff either way.
Shaw, I actually do recommend a sort of joystick for sketchup and google earth if you're using them a lot (which I haven't lately). For the price it is quite good (~$80).
With a bit of practice you'll be navigating and orienting in 3d with your left hand and using your regular mouse normally with your right (or vice versa if you're a lefty).
Yeah the boom was certainly meant as a complement, though I really meant your were the "bomb". Oh well. Maybe I'll get motivated this weekend and create a 3D histo in Sketchup (my almost most favorite program). We'll see.
I read your article on these 3D visuals you made and I am fascinated in how you've created them. I have been trying to recreate your methodology for your 3D Scatterplot with Axis Dashboard with my own data, sadly I have been working on it for 2 weeks and no luck. Here are some of my main questions:
- I have to say I am rather confused on understanding which axis is which; for example on your Surface 1, Axis4_grid1 is labeled 0.03 and Axis4_grid1 is labeled 0.00 which really has threw me for a loop.
- I'd also love to know how you went about picking your parameter values for X top/bottom, Y top/bottom, and Z top/bottom.
- If I wanted to make the X and Z axes interchangeable via a parameter would that effect the parameter values for X top/bottom, Y top/bottom, and Z top/bottom (with respect to my data, once I understand how you selected those values)
- I followed your SQL query and was perplexed on your choice to use 0.33 and 0.66 to alter your x/y/z values; what is the significance? is it related to the fact there are to be 3 planes?
- And lastly, where is this data from? what is it/where did it come from?
To others, here is Bora's article: Going 3D with Tableau | Bora Beran
I really hope to hear back from you on this. I understand it's been some years since you've created these, but your help would be amazing.
PS. I will also be leaving this as a comment on your article posting:)