Skip to content

Please stop me before I barf again

Pointing to some horrible graphs, Kaiser writes, “The Earth Institute needs a graphics adviser.”

I agree. The graphs are corporate standard, neither pretty or innovative enough to qualify as infographics, not informational enough to be good statistical data displays.

Some examples include the above exploding pie chart, which, as Kaiser notes, is not merely ugly and ridiculously difficult to read (given that it is conveying only nine data points) but also invites suspicion of its numbers, and pages and pages of graphs that could be better compressed into a compact displays (see pages 25-65 of the report). Yes, this is all better than tables of numbers, but I don’t see that much thought went into displaying patterns of information or telling a story. It’s more graph-as-data-dump.

To be fair, the report does have some a clean scatterplot (on page 65). But, overall, the graphs are not well-integrated with the messages in the text.

I feel a little bit bad about this, because I’m involved with the Earth Institute. I should be their graphics adviser! I’m actually surprised they didn’t ask me for advice on this. I gave a talk to the Earth Institute postdocs a year or so ago, so they should know I like graphs!

P.S. Nathan Yau reproduces the report’s Figure 11, which is just as bad as the exploding pie chart shown above.


  1. […] someone please get Andrew Gelman a barf bag before he makes another mess. AKPC_IDS += […]

  2. Jon Goldhill says:

    Agreed. Given that the numbers add up to 99%, it seems sad that effectively nobody has sufficient ‘health’ plus ‘education’ (for example). I dread to think what questions were asked.

  3. Rick in Chicago says:

    My first observation is that the pie wedges were all about the same size, which suggests that the question or the answers or both were pretty much meaningless to the respondents. I have no idea what “sufficiency” is, and I suspect that neither did the respondents. So they picked one at random (which with randomized ordering of the responses would be very easy to do.) This assumes that the question was simply “Which of these things defines sufficiency to you?” or something very similar.
    If the result was after some clustering of respondents or factor analysis, then I’d call the results a statistical FAIL, in that the underlying items lacked sufficient discriminating power. (This is the multivariate version of my first answer.)

  4. Nick Cox says:

    If you look at the report, you will see that the pie chart is nothing so simple as the distribution of answers to a single question. As I understand it, it is a summary of the relative importance of different dimensions in a composite index. The context is needed.

    That said, the report like many others mixes quite different graphical styles and the graphs span a range from good to grotesque.

  5. Jonathan says:

    I got a pretty terrible one from Bloomberg.

    I was so furious I actually tried to replicate the opinion stuff, and can’t get the exact same measure (probably due to my lack of knowledge of the GSS). I encourage someone to try.