Update 2, 6.6.14: Due to the criticism of this study on the web, I have reanalyzed a part of this study. See my comments here.
We should by now know that that which we call a rose; by any other name would smell as sweet; but it seems we still haven’t quite internalized the message. Proof comes from a new study in PNAS that looks at the correlation between hurricane names and death tolls. The authors around Kiju Jung find that:
Feminine-named hurricanes (vs. masculine-named hurricanes) cause significantly more deaths, apparently because they lead to lower perceived risk and consequently less preparedness. Using names such as Eloise or Charlie for referencing hurricanes has been thought by meteorologists to enhance the clarity and recall of storm information. We show that this practice also taps into well-developed and widely held gender stereotypes, with potentially deadly consequences.
It sounds a bit crazy that the perceived femininity of the name should create effects as large as those reported in the paper, but on a first glance the methods look rather convincing [update: Well, see below]. If correct in its conclusions, the study certainly has some important implications for the communication of hazards in general.
What I specially liked from the methodological perspective is that the authors first analyze the available public data on death tolls in hurricanes, which shows the correlation with the names, and then run further experiments to show that the detected correlation is actually causal. I guess there is still some room for discussing potential untreated confounding effects, but all in all it looks like a great example for students to think about confounding effects, and how to get from an observed correlation to some confidence about causality, running further experiments and so on. Quite sure I’ll have my students discuss this one in future research design courses.
Update 5.6.14: OK, I might have to reconsider whether this is really a good example for students, although it might even get better by that: Bob and GrrlScientist claim that the model is not correctly specified and that a suggested correction removes the effect of femininity.
Bob’s argument looks quite convincing as well. Man, I don’t know, I would have really given this one the benefit of the doubt, assuming that both the authors and the reviewers should be able to pull off a more or less correct linear regression for a paper in PNAS. I mean, I can sort of understand this, although I also think the reviewers should have caught it, but checking the residuals in a linear regression with 90 data points ???. OK, what additional insight do we have: 1) good that their data is available 2) one has to check everything 3) it seems that journals have to find volunteers or pay statisticians to REDO the stats for any, at least for any major paper that is published. It’s costly, but once a paper is published with errors in the analysis, correction is difficult and collateral damage seems unavoidable.
Jung, K.; Shavitt, S.; Viswanathan, M. & Hilbe, J. M. (2014) Female hurricanes are deadlier than male hurricanes. PNAS, in press.