There is this trend that seems to be popping up everywhere of using radar charts such as this one:
For the record people, these are bad. They imply that there is some circular relationship about your data points. They are line graphs made into a circle. If you want to indicate the volume under a curve, make a line graph, fill the area under your curve, and then consider it done. Don't use these unless your data points progress from A → B → C and then back to A.
Please?
EDIT: I should add - don't use line graphs for things that don't progress from A to B to C either. For those things, use bar graphs, or column graphs.
One alternative to radar charts that seems to work pretty well under certain circumstances is the parallel coordinate graph: http://www.b-eye-network.com/view/3355
If nothing else it allows you to compare the values for different entities, which is practically impossible in a radar chart.
That's interesting. Very good for finding exceptions to rules. It has a similar problem that the slopes of the lines look like they are relevant when they're not, but I can live with that, for the gain of the picture view. The interactive concept is cool too. I'll bet in about five years we will have interactive graphs on your average pdf & excel chart. It shouldn't be long now.
Parallel coordinate plots can be a useful way to take stock of a high-dimensional data set, but as information visualizations I think they're kind of misleading. On that linked page describing them, the author says: "We must be careful to read nothing of significance into the slope of each line segment or the overall pattern formed by the line as a whole. The slopes and overall pattern would look completely different if I rearranged the order of the variables."
And therein lies the problem. The intuitive way to read a line plot of connected points is wrong. So while I think PC plots can be useful for your own perusal (as an analyst, say), I would think they would mislead at first glance (and maybe upon subsequent glances, too, unless the glancer understands what high-dimensional data sets are and why you might want to examine them this way).
Post new comment