Roland LÃ¶ÃŸlein, a media student at University of Applied Sciences in Augsburg, presents meteorological time series data in 3-D in a class project called Synoptic. Rotate and zoom in and out on the different time lines, select different metrics, and compare against the corresponding time series on the bottom. After a few minutes of playing with it, I’m still trying to decide whether or not it’s useful, but I think it’s more of an experience than it is an analytical tool. It’s almost like exploring a map, but instead of rolling hills, you get dips and peaks in a chart. Interaction is smooth and the visualization scores well in aesthetics.
I hate 3d graphs. They are worse than useless– they distort comparison through perspective and disallow simultaneous viewing of all data.
They are for games– not charts.
I can see where 3D could be useful for exploration–you certainly get a different perspective of the data when you can fly around it.
But I think it might best be used as an entry point to further views. In this Synoptic demo, I would, like to be able mark a few curves that look interesting, then have them output into a 2D chart that I can see more clearly, and perhaps actually use/print-out/etc.
I disagree. Yes you get a different views as you fly around because each view is a distorted snapshot with the rest of the data occluded.
The graphic shown requires 2 dimensions plus colour– if the 3rd dimension (y) here is a real scale then the colour can be in shades of grey- with a scaled color ramp by the side.
If a further dimension is required then you can plot separate panels on top of each other– it’s a timecourse so that’s easily interpretable.
If a 5th dimension is required then you should start thinking about plotting derivatives of the data.
3d or interactive graphics have a place of course and that place is when you are interested in trees, networks, and the like. In such cases you are interested in the pursuing (exploring if you like) the pathways and connections between entities and not an informative summary of scale.
3-D is useless a lot of the time, but sometimes can be pretty useful. For example…
Regardless of whether you can rotate or tilt the data, sometimes that literal third dimension leads to better understanding. I do health economics and a consistent finding from cross section data is that health expenses for individuals in one year are strongly correlated with expenses in a subsequent year. Many years ago I had a file with data from Medicare patients. I plotted year one expenses on the y axis; year two expenses on the x; and the number of patients with any particular x,y combination on the (vertical) z axis.
My priors were that I should perceive some sort of ridge confirming the correlation. Instead, the bulk of the data were relatively tightly packed near the origin. Because all of the data were in the first quadrant the correlation (slope of x on y) pretty much was guaranteed to be positive. Because the vast bulk of the data clustered near the origin, they generated a high r-squared. The picture was not a ridge but a mainly symmetric ski-slope from the origin outward. I needed that third dimension to get the story right.
Pingback: Michael Romanowicz: Digital Creativity & Experience