Also this week: with
four one free parameter s I can fit an elephant, good news about lack of racial or gender bias in NIH grant reviews, social psychology continues to not replicate, speciation beer, and more
Two more canonical social psychology results (both cited over 1300 times) failed to replicate in big preregistered multilab replications. In both studies the treatment effect is precisely estimated to be very near to, and not significantly different from, zero. Here’s a reasonably comprehensive list of all replications from cognitive and social psychology. (ht @noahpinion)
Relatedly, this is pretty funny. 🙂
Good news: a new unreviewed preprint finds no evidence of racial or gender bias in NIH grant reviews, not even if you restrict attention to white male reviewers. The estimate is statistically precise, so the finding of no effect is not a matter of lack of statistical power. The result comes from a clever experiment involving assigning fake names to real NIH grant proposals. I’ve skimmed it, it’s from researchers who work on this topic for a living, and the design and analysis were preregistered and look very solid and thorough to me. For instance, the analysis considered both the reviewers’ numerical scores and their word choices in their reviews. And there were procedures to detect reviewers who may have recognized the PIs as fake, and to assess the sensitivity of the results to those reviewers. As the authors note, the study addresses only how study sections evaluate proposals, not factors that might affect the proposals themselves (e.g., variation among PIs in access to mentoring and other support). And it doesn’t address other possible reviewer biases, such as favoring proposals by famous PIs. (ht @noahpinion)
An overview of how Columbus, Ohio is bucking the decline of many northern US cities, thanks in part to Ohio State University.
Individual vs. group selection, California politics edition.
This is cool: an as-yet-unreviewed preprint describing an equation with one free parameter that can fit any bivariate data plot arbitrarily well. Turns out Von Neumann was wrong: you only need one free parameter to fit an elephant, not four. I will have to add this line to my list of the best sentences scientists get to say:
This single parameter model provides a large improvement over the prior state of the art in fitting an elephant.
The secret sauce is the discrete logistic equation (aka the logistic map) in the chaotic regime. I’m not sure this has any direct practical applications, but the conceptual implications seem pretty profound: the usual intuition that more free parameters = more flexible model is wrong. Analogous to how the logistic map shows that even simple ecology can generate complex population dynamics. (ht Marginal Revolution)
A new book about efforts to use randomized controlled experiments in areas of social science in which that’s never been a common approach. Looks interesting; this is a debate I follow a bit on blogs. (ht Marginal Revolution)
An interview with prize winning philosopher of science Michela Massimi (ht @RELenski). Here’s one good bit:
Dismissive claims by famous physicists that philosophy is either a useless intellectual exercise, or not on a par with physics because of being incapable of progress, seem to start from the false assumption that philosophy has to be of use for scientists or is of no use at all.
But all that matters is that it be of some use. We would not assess the intellectual value of Roman history in terms of how useful it might be to the Romans themselves…I see the target beneficiary [of philosophy of science] as humankind.
Speciation beer. (ht Andrew Hendry, via Twitter) In the past, they’ve also offered Genetic Drift beer, Cambrian Explosion beer, etc. Even though I am into evolutionary biology and am a beer nerd, I confess that this brewery’s distribution method wouldn’t really be my thing even if I lived close enough to their brewery to participate in it, which I don’t. But Meghan lives in Michigan, and right now she’s on sabbatical and so presumably has tons of time on her hands. So anybody who wants a bottle of this, email Meghan your order and I’m sure she’ll happily oblige. 😉 In the comments, please suggest ecological and evolutionary jargon terms that would be good beer names. 🙂 I’ll start: “Coexistence” would be a good beer name, especially for some hybrid style.
A lovely illustration of why the academic PhD advisor-advisee relationship is fraught even with the best advisers (power-imbalance) yet can be wonderful (from Math With Bad Drawings). Basically PhD Comics but with a heart-warming twist.