Because every place worth going can be found on a topo map.
Like many things, this blog has a purpose, though it is a purpose that is only secondary to its point. To whit:
It is perhaps worth adopting as a rule of thumb that if a data visualization takes ten pages of text to explain it’s not doing its job. This paper has some really fascinating ideas about how to display complex data in a way that allows the viewer to preform the sorts of visual analysis of multiple demensions (specifically intensity and variety) simultaneously. Unfortunately, what they end up with is this:
Wha?
Now, that’s a little ingenuous of me. Having read the article I see what the authors were trying to do and I can get an interesting sense of the data out of this graph. That said, in a far shorter amount of time I could have thrown the raw data into Stata and come out with some basic cluster analysis which would have told me more.
This graph is confused, the glyphs are too big and oddly shaped (in fact, the size of the glyphs is scaled to the height of the player. Can you tell? I can’t.), the axis aren’t marked (and marking them would make the graph even more illegible), the three dimensionality is impossible to decipher, the coloring is completely redundant with the axis, and the RGB color combinations that represent relative distances along each axis are non-intuitive for anyone who hasn’t spent a great deal of time with a color wheel.
The glyphs should be simpler, area shouldn’t be used to represent a linear quantity like height (it would be more intuitive to represent weight with the size of the glyph anyway), three dimensions cannot be well represented in two, and combining multiple color spectrum at once is just too confusing. An addition of some form of interactivity, perhaps allowing users to navigate through the three dimensional space, or at least rotate around it, while simultaneously getting rid of the chart junk glyphs and the confusing colors might make this system somewhat workable. However, as is, the entire endeavor is actually quite embarrassing.
Compare, if you would, to this article, whose centerpiece is this:
On cursory inspection, this appears as, if not more, inscrutable than the first chart. Examine it for a minute, though, and its internal logic start to make sense, even without having read the article or knowing what it is supposed to be representing (though the scores quickly give it away as a tennis match.)
The layered tree structure is almost instantly recognizable, and the color coding of the player names at the top gives a good visual clue as to what the colors mean throughout the chart. The winner of each point, game, set and match gets all subsidiary nodes shaded his color. The scores, highly legible, provide context and order, and the magnifying windows provide at least a modicum of interactivity. The wealth of accessible data in the chart is somewhat overwhelming. Each racket hit is recorded and mapped and can be seen in magnification in the little rectangles inside each point. That is a huge amount of discrete data being visually displayed all at once.
A major problem with this chart is that its simply ugly. Though it can be navigated, and eventually understood without too much effort, it does not invite the viewer to give it a chance. A complex chart, like a good lecture, starts with some element that the audience already has a good grasp on. Maybe that’s a prominent aggregate trend line used as a comparison for more complicated breakdowns. Maybe it’s an inviting slider bar at the bottom that encourages manipulation of the display in a simple way that will be enlightening. In short, the issue with this graph is it doesn’t have a clear beginning.
Clearly, both of these examples show that representing multiple kinds of data on the same graph is a difficult proposition. Even when done relatively intelligently, as in the second paper, it can leave the viewer grasping at straws.
I’m also supposed to review a chapter of Few’s book on data visualization, but I still have not got my hands on it. Looks like I’m building up a bit of a backlog.
