A beautiful example of a time-series plot outdoing a bar graph

There’s really no need to link to Junk Charts anymore, since if you’re interested in data display, you’re probably reading it already . . . but here’s a nice one:

The bad graph from the newspaper:

fashion.gif

The hypothetical improvement:

redogarment.png

I think the biggest marginal improvments can come from turning tables into graphs, but in this case, the better graph actually takes less effort to make as well as taking less space and being clearer (and also suggesting the potential for further improvments such as including data from other countries).

I’d probably connect the dots rather than draw the regression lines. Then the big dots wouldn’t be needed. I’d also label the countries on the right side of the graph rather than below the lines–that would remove the need for different colors and would allow room on the graph for the E.U. and Japan, for example. Also I’d label the y-axis at $0, $100 billion, and $200 billion.

2 thoughts on “A beautiful example of a time-series plot outdoing a bar graph

  1. Yeah, simple time-series plots are generally much better than barcharts–but in some cases a time-series plot still doesn't tell as full a story as it might.

    Compare these two time-series plots. One is pretty standard, the other is a variant that sometimes works well when one is comparing only two series (discretionary spending is chosen because it reflects each administration's priorities better than spending that includes entitlements):

    A standard time-series approach

    An alternative presentation of the same data

    I think the alternative presentation makes it easier to see how each administration has viewed its overall priorities, the total of the two components, and comparative growth.

    I don't know about you, but in the standard plot my eyes run over to the axes more frequently.

  2. That second presentation is nice. I often have to deal with datasets that split naturally into "good" and "bad" populations (credit scoring), and I tend to use ROC curves rather than KS curve presentations for precisely the same reason.

Comments are closed.