Gary Smith gives a clear non-technical explanation of why not to take that himmicanes study seriously. Further background here.
Gary Smith gives a clear non-technical explanation of why not to take that himmicanes study seriously. Further background here.
Sure, the overall relation doesn’t hold. But, if you look at the subgroup of hurricanes with stripper names, the effect is highly significant (for male or female strippers).
Steve:
Good point. It’s Hurricane Butch vs. Hurricane Trixie here.
But what if you break it down further by trans or cis gender?
WTF folks?
This is the sort of conversation that makes the blog feel hostile to women and trans folks. It’s the sort of thing that was roundly criticized in the comments thread on the Bisexual PNAS post.
Anon:
The above comments are mocking the himmicanes paper which is an example of how gender essentialism can lead to bad science. See the linked criticism by Gary Smith for further discussion of this point. The social-science and statistical issues here are similar to those that arose in the beauty-and-sex-ratio papers that we discussed many years ago (for example here).
My apologies, Anon, if my comment came across as hostile to women and trans folks. That was far from my intent. My comment arose from my (scientific) tendency to see things from a finer detail than often first meets the eye. In particular, as an older woman, I see my socialization and experiences as a girl in the 1940’s and 1950’s as having had a strong effect on me — and that is something which a trans woman my age would not have experienced, just as a cis woman a generation or two younger than me would not have had those same experiences.
“Ransacking data rarely ends well.” – I’ll be using this phrase liberally for the rest of my career…
“prestigious journals sometimes publish rubbish. Consider, for example, […] the Lancet’s publication of a paper claiming that Jews can postpone death until after Passover.
The hurricane paper is almost as bad.”
Does anyone have an (informed?) opinion on that one?
I have written several papers on the postpone-death literature:
https://www.garysmithn.com/papers
Gary:
What a cool set of papers! It’s like a sane version of Freakonomics. But some of the links to the papers don’t work. You should fix them.
They also conducted a series of lab / online experiments, with university students or mechanical turks users, respectively.
The experimental condition was the name of the presented hurricane. The dependent variables were responses to hurricane warnings.
The experiments are part of that paper.
Norris:
Yeah, I guess they should’ve done a better job of the analysis of the historical hurricane data (see the criticisms in the Gary Smith post and elsewhere), also reported their experimental results, and then retitled their paper, “No evidence that female hurricanes are more or less deadly than male hurricanes.” In their abstract, they could’ve written something like:
What got the paper attention was not the lab studies but the claim about real hurricanes. And that’s the claim that did not hold up under scrutiny.
P.S. Had the authors followed my advice and titled their paper, “No evidence that female hurricanes are more or less deadly than male hurricanes,” then I expect it would’ve been much harder to get it published in PNAS. This is an example of selection bias: the wacky unsupported claim gets published (with the assistance of the lords of National Academy of Sciences) and treated as fact; a sensible data-based conclusion could’ve been published in some low-grade place and not received all the adoring media attention.