Himmicanes and hurricanes update

Stuart Buck points us to this new paper by Gary Smith that eviscerates the notorious himmicanes and hurricanes paper. Here’s how Smith’s paper begins:

Abstract
It has been argued that female-named hurricanes are deadlier because people do not take them seriously. However, this conclusion is based on a questionable statistical analysis of a narrowly defined data set. The reported relationship is not robust in that it is not confirmed by a straightforward analysis of more inclusive data or different data.

Keywords
Hurricanes; Data grubbing; Sexism

Ha ha. I’m just bummed he didn’t use the term “himmicane,” which I think treats this topic with the seriousness it deserves.

To the best of my knowledge, the authors of that joke paper, and the editors at the journal that published it (PPNAS), have refused to acknowledge the flaws in the paper, nor have they apologized for wasting all of our time with it. I probably should not be surprised by this never-back-down attitude. But I don’t have to like it.

16 thoughts on “Himmicanes and hurricanes update

  1. >wasting all of our time with it

    There’s no rule which obligates anyone to read every article in PNAS, much less continually blog about it. If you choose spend time dwelling on this paper, that’s on you.

    • Jacob:

      I never would’ve heard about the original himmicanes paper except it was covered by the national media.

      And, I think it’s an embarrassment for PPNAS to have published this paper. If I publish a bad paper, I apologize and try my best to correct the record. I don’t leave the paper up there with the comment that nobody is forced to read it.

      It’s “on me” that I continue to write about the himmicanes paper, but it’s “on PPNAS” that they published it and continue implicitly to stand by it.

  2. ST:

    I’m not saying PPNAS should erase the memory of this paper. They can keep it up on their website. But I think it would be appropriate for them to link to criticism, including this new paper by Smith and (ideally) a note by the original authors acknowledging their errors.

  3. Looked at in a particular way, this is not a joke paper. It is one exchange in a very long political argument, and it is not to be judged by its consistency or plausibility, but on whether it adheres to, and successfully propagates, a particular political view. From this point of view, contributions such as this paper are not validated (or not) by statistics. Instead, statistics is validated (or not) by its track record of being politically useful (which is actually quite good, since its roots are in describing the situation of the poor). What some people require is not evidence based policy, but policy based evidence.

    While I do not adhere to this point of view, it would be naive not to recognize that it exists.

    • I wonder if Andrew McD could give his definition of “require” (as in “What some people require is not evidence based policy, but policy based evidence.”) Is it perhaps just a synonym for “want”? Or if not, what?
      And then there’s always the question, “*Which* people require …?”

      • I didn’t explicitly decide between those two words. Perhaps I am influenced by dealing with requirements in the context of providing computer systems, but to me a want is less likely to be acted upon than a requirement, and may simply be a description of a state of mind.

        There is background to the “policy based evidence” phrase at https://en.wikipedia.org/wiki/Policy-based_evidence_making and these allegations, if true, fit the model either of a policy-maker explicitly asking a researcher for research in support of a particular policy, or of a researcher searching for evidence in favor of a particular policy, in the expectation that this will increase the likelihood of them receiving further funding.

        (Apologies for the writing style. Other comments suggest to me that this might be causing problems, but do not provide specific enough information for me to attempt to improve it. Indeed, they are general enough that, were they capable of producing improvement, I could have generated them myself and benefited from the improvement before posting the comment).

        • I understood you right away.

          Basically some people want to believe that females and feminine things in general are taken less seriously, and they’re just looking for ammunition in support of this theory so they can use it. The process is essentially “search through all knowledge until you find results that support your conclusions, and promote those” rather than “check against reality to see if these conclusions make sense”

          The same is true for lots of politically controversial topics. Some people are supporters of a belief independent of whether there is evidence against that belief, or whether there is a mixture of confusing evidence.

        • Thanks for the reply. My view might be expressed as “some people want policy-based evidence, and therefore require it”.

          Also to clarify (my use of the word want) further, I think of “what someone wants” as often influencing their interpretation of some concept — e.g., I believe that one of the problems people have in understanding p-values is that they want (or at least expect) the simpler concept “probability of something given the data” rather than the more complex concept that is needed to accurately understand what the p-value is (and isn’t). In teaching statistics, I often contrast “what we want” with “what we get”, in the hope that it will help students realize that they are not getting what they really would like to have.

  4. I submitted a response to the Himmicames paper showing it was crap (plot the residuals, folks!) but the editors didn’t send it to review because showing a study is wrong doesn’t advance the study. Even getting that a mission out of them was like getting blood out of a stone.

  5. I love you Gelman, but if a college student turned in a paper with that high of a quote to original text ratio he would almost certainly get an F (referring to your WP piece).

Leave a Reply to Keith O'Rourke Cancel reply

Your email address will not be published. Required fields are marked *