Skip to content

Florence Nightingale’s graph

Chris Zorn pointed me to this news article by Julie Rehmeyer.

Given the context, the graph is impressive and important. But given what we know today, it would’ve been even better as a line plot. (Rehmeyer suggests a bar graph but I assume that’s just because she doesn’t know about line graphs; see here for a simple example.)


  1. chewxy says:

    Hard to believe she drew that without computers -_-

  2. Nick Cox says:

    As you say, this would be much better as a line graph. So why doesn't everyone say that?

    It's curious that this graph repeatedly gets largely uncritical reports like that quoted here — as if the story were new and as if the graph were not intensely problematic. The Economist, not usually known for sentimental slush, went all gooey about it in its Christmas issue for 2007.

    My guess is that critical senses are dulled because of respect for Nightingale's other achievements (and her status as an icon of various kinds) and because it's, in journalistic or historical senses, part of a cracking good story.

    A very small detail, but again one that often arises, is that Nightingale didn't intend to call the _design_ a coxcomb; that was an offhand comment she made about the graph.

    A central point is to me that the circular format is at best justifiable if the pattern shown is strongly seasonal. (Even then it is far from obvious that circular format hinders more than it helps.) But we have two complete years, which is not much to go on. Even gross seasonality (e.g. peak months) is quite different in each year, and that seasonality is less obvious than secular variations — the fine structure of which presumably reflects variations in casualties, changes in nursing care, outbreaks of contagious infections, etc.

    To make a further standard point: there is dimensional ambiguity when sectors (strictly annuli) are shown. Are the widths or the areas shown to scale? People see areas, but they don't perceive them very accurately. Even in geology where rose diagrams are popular, and the audience quite numerate, it is common to find graphs presented which may or may not be dimensionally correct as it's not clear which rule was used to produce them.

    Does any one know a source for the original data (i.e. numbers)? Although Nightingale is much reprinted and anthologised, the original works relevant to this graph appear very elusive outside of extremely good libraries. I gather that a reissue is likely soon in the project from Wilfrid Laurier University Press.

  3. Andrew Gelman says:

    Nick: I don't know the answer to the question you pose in your first paragraph, but from reading Howard Wainer's books, I've learned that simple scatterplots and lineplots are relatively new inventions. For some reason, these methods, so obvious to us, weren't so obvious to people in the past. So I guess the people now who don't notice that Nightingale could've used a lineplot are the same as the people now who use barcharts and piecharts etc.

  4. Nick Cox says:

    My suggestion is mostly that the reason that people are so uncritical of this graph by Florence Nightingale is that it was by Florence Nightingale. It might seem ungracious, if nothing else, to be carping about the graph given all the context and associations; or petty to be critical given the presumption that Nightingale's work really made a great deal of difference to the world.

    I don't think the history of statistical graphics, an avocation of mine, helps much here, except to deepen the puzzle. Lineplots predate anything based on subdivisions of the circle by several centuries. Pie charts and variations on rose charts seem to have arisen around 1800. Line plots go back to medieval times. But then Florence Nightingale might not have been especially unusual if she stuck to a relatively original design that occurred to her, rather than something simpler and more commonly used. Most people using statistical graphics do that much of the time, myself included.

  5. A. Goodman says:

    You people are missing a key point. This format for the graph stacks common months on top of each other beautifully, which is harder to accomplish as naturally in other graphical formats.

  6. Andrew Gelman says:

    A: I think you could do the comparisons by month better by overlaying several years with lineplots. The circular structure of the above plot is indeed beautiful but I don't think it conveys the information very well. To use some language we've been using a lot lately, the Nightingale graph is excellent at "data visualization"–it is attractive, grabs one's attention, and pushes one to think about the data–but it's not so great as "statistical graphics" in that it does not directly facilitate a deeper understanding of the data.

  7. Nick Cox says:

    I don't see what point A. Goodman thinks is being missed.

    As above, the circular scale is not a good idea because seasonality is not obviously the main story, and even if it were, it's more difficult to compare years with two circular plots than if other formats.

    I wouldn't want to deny that this is an attractive graph, but that is a pretty weak claim.