Friday links: why philosophy of science, evolutionary biologist vs. divining rods, and more

Also this week: the Trump administration’s latest attacks on, and neglect of, expertise and information, and how to push back against them. Which is pretty depressing, so there’s an Abba vs. Unix link to make up for it. Also, new experimental data on single vs. double-blind peer review, what to get Stephen Heard for Christmas, and more.

From Jeremy:

The theory of empty theories. 🙂 Alternative title: a conceptual/graphical model of untestable conceptual/graphical models. Semi-related old post. (ht @kjhealy)

Evolutionary biologist Sally Le Page on how 10 of 12 UK water companies still use DIVINING RODS to locate underground water pipes. Yes really. Northern Ireland Water and Wessex Water are the only UK water companies that don’t use divination. I now want to book them to replace the water pipe to my house here in Canada, just to thank them for not living in the 17th century.

Why philosophy of science can be very valuable for an undergraduate science education–and why undergraduates often don’t realize this. I’m tempted to suggest that managers from every UK water company except Northern Ireland Water and Wessex Water should be required to take a philosophy of science class. So should everybody who filled Sally Le Page’s Twitter mentions with the argument that divination must work because water companies still use it. But honestly I doubt it would do much good. Related: if you want to dip a toe into philosophy of science, here are our suggestions on where to start.

Advice from a former higher education lobbyist on how to lobby against the graduate school tuition waiver tax in the House budget proposal. And how not to.

The Trump administration is considering political scientist Thomas Brunell for the top operational position at the US Census Bureau. Brunell often testifies on behalf of Republican redistricting efforts and wrote a book arguing that competitive elections are bad for America. His nomination and confirmation would be a break with the longstanding and essential practice of filling the position with a nonpartisan career civil servant with statistical expertise. I suggest that if Brunell is nominated, it’s worth contacting your representatives to oppose him. The government should make political decisions based on accurate information, not politicize the gathering of information. Politicized information isn’t information. As scientists, we have even more stake than most citizens in making sure the government cares about information.

The White House Office of Science and Technology Policy still has no leader, the longest its ever gone without one since it was created in 1976. No one’s been nominated. And it has about 1/3 of the staff it had under the Obama administration. And no mandate to do anything.

Now I know what to get Stephen Heard for Christmas. Actually it’s a bit pricey for me. You’ll have to ask Santa, Stephen. 🙂

And finally: Abba vs. Unix 🙂 Gimme gimme gimme a man Unix shell command after midnight. (ht @kjhealy)

From Brian:

A new paper with arguably the best experimental design yet on single vs. double blind peer review was published. It was for abstracts for a data mining conference. Every abstract was read by two reviewers single blind (knew the authors) and by two reviewers double blind (didn’t know authors). On gender the study found: “The influence of author gender on bidding or reviewing behavior is not statistically significant [p=0.16]. However, the estimated effect size for Wom [coding for female] is nonnegligible.” This is very consistent with what I’ve seen in other less ideal experimental designs. There is no killer conclusion of gender bias, but there is almost always a low grade signal which is cause for worry. The whopping effects found were bias towards famous researchers and researchers who worked at famous institutions. A rather conclusive result there I thought. They did not find a bias towards US authors or authors from the same country as the reviewer nor towards industry vs academic. They did not study seniority or career stage or reputation aside from a fairly restrictive famous/non-famous category.

8 thoughts on “Friday links: why philosophy of science, evolutionary biologist vs. divining rods, and more

  1. Brian, nice find with that experiment. Like you, I’m unsurprised by the results.

    I’m curious: are abstracts for computer science conferences mostly sole-authored? If not, how did the investigators decide whether co-authored papers were by “famous” researchers. Imagine that a grad student is first author on a paper with their famous supervisor as the last author–is that a paper by a “famous” researcher?

    • If any author was “famous” the paper was coded as famous. The definition of famous was based on numbers of papers published an abstracts presented but it was restrictive enough to be only the top 57 people in the field. Similarly if any author worked at a famous university (=top 50 in compsci) or famous company (=top 4) it was coded as famous institution.

      I really wish they had just put in a career stage or # of papers published or such, because I bet that seniority had a real effect too, but they didn’t measure that.

      Just noticed that they linked to a pre-print where they did a meta-analysis combining their study with several other studies. At which point the bias against women is statistically significant. I kept waiting for a meta-analysis. I saw so many studies where there was a trend of gender bias but not statistically significant. I was sure if somebody aggregated them up, they would become significant which appears true. That said the gender bias is smaller than the famous person bias.

      • And I’m sure fame (or seniority) is collinear with gender–senior academics include a higher proportion of men than junior academics. Does the study, or the associated meta-analysis, try to separate them?

      • They did multivariate logistic regression (i.e. all the variables in the same regression). That addresses some of the issues of collinearity and creates others. But as the meta-analysis showed their estimates of effect sizes on both gender and fame were very in line with other previously published estimates (not all of which studied the same set of predictor variables as they did). So I doubt they had severe problems from the collinearity.

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.