Also this week: YouTube vs. Evolution 2016, satirical psychology paper, does moving to Harvard increase your scientific impact, the origin story of BEF research, all your fMRI data are
belong to us false positives, and more.
Some of the talks at this year’s Evolution meeting were recorded and posted to youtube. A list is available here.
There was a column in the NY Times regarding a new study of economists that found that policies that allow new parents to extend the tenure clock have strong negative impacts on women. The argument is that many men use this time to do more work, whereas women cannot do so (because, as the piece says, birth is not a gender-neutral event). I’ve talked with a lot of people about this general issue, and it seems that everyone knows men who seemed to use that time to get more work done. But there are also plenty of men who use the time to be engaged parents, and my life was certainly improved by my husband getting a release from teaching when we had our children. (Also: I was shocked at how low the tenure rates were in the piece — 30%?!) UMich now has a new policy that allows birth mothers two semesters of modified duties after the birth of a child. I think this is a good thing, but also will be interested in seeing how it plays out. A concern I have is that it will seem like women had extra time, and so the bar will be raised. I very much hope that is not the case! (Jeremy adds: Meg beat me to this link by literally one minute! I second her comments. Here’s the underlying (unreviewed) preprint, which I haven’t read and can’t vouch for.) More from Meg: I wrote the previous blurb a week ago. In the meantime, there’s been a lot of discussion of that column. Here’s one important piece critiquing it, entitled “Don’t change your family friendly tenure extension policy just yet“. It points out that the full study found “no evidence that gender-neutral tenure clock stopping policies reduce the fraction of women who ultimately get tenure somewhere”. (The 30% number relates to tenure at the first institution at which someone was hired.) It also notes that there are so few women hired in economics in each year that there is a lot of noise in the baseline estimates. If you read the NYTimes piece, it’s worth reading the second one, too. (More from Jeremy: aha! I thought that 30% number sounded way too low. It’s of course very common for people to eventually get tenure someplace other than where they were first hired, as Brian and Meg themselves illustrate. So I agree with Meg: that the headline result is to do with “rate of getting tenured wherever you were first hired” as opposed to “rate of getting tenured” affects the interpretation. And I found that second link from Meg very helpful. To my mind it raises serious statistical concerns about the original study. But I’m not an expert on this topic so perhaps I’m still missing something here. Anyway, it’s my impression as an outsider that this sort of criticism of unreviewed preprints is pretty normal in economics; the criticisms often lead to pre-submission revisions. Thanks very much to Meg for the update, good catch!)
A disaster for fMRI researchers, but good fodder for an intro biostats course: if you construct fake fMRI experiments by randomly assigning data from healthy control individuals to both the “control” and “treatment” groups and then run them through the most popular analysis packages, you get false positives up to 70% of the time. The issue is the faulty way popular analytical packages deal with autocorrelation in fMRI data. (ht @kjhealy)
If you’re like me, you probably think that there’s been a long-term trend in the US for the large majority of colleges and universities to cut back on full-time tenure track faculty and replace them with part-time adjuncts. If so, you’re wrong (summary here, and see the comments for a deep dive into the details of the data). It’s true that the percentage of faculty who are full-time has dropped a lot. But the absolute number of full-time faculty in the US is actually increasing, and indeed has grown faster since 1990 than pre-1990. That growth parallels the growth in the student body; the ratio of full-time faculty to students (~1:25) has remained incredibly stable for 40 years. Part-time adjuncts remain a minority and their usage is only growing slowly at most types of institutions, with two big exceptions: two-year colleges and for-profit colleges. And before you ask, no, focusing on full-time vs. part-time faculty does not obscure a massive decline in tenured and tenure-track faculty. Not all full-timers are tenured or tenure-track, of course. But the percentage of full-time faculty who are tenured or tenure-track is down only about 5 percentage points since the early 90s, and depending on type of institution has been flat or even increasing for the last decade. Bottom line: the boom in part-time adjuncts in the US mostly reflects the recent boom in number and size of for-profit colleges (a boom that now seems to be reversing), and the previous boom in 2-year colleges. It has not come at the expense of full-time faculty at 4-year institutions, tenure-track or otherwise. Note that I don’t think you should leap to any conclusions about the overall state of the academic job market solely from these data. They’re just part of the picture. (ht @dandrezner)
Andrew Gelman and Jonathan Falk have written a satire of the sort of psychological studies Gelman often critiques on his blog.
This is a link from one of our commenters a while ago, that I forgot to highlight at the time. I think it’ll still be of interest to many of you. Deville et al. 2014 asked what effect if any institutional affiliation has on scientific impact. Put crudely, if you move from Small Obscure U to Big Famous U, does your impact increase? Whether because people just pay more attention to papers from Big Famous U faculty, or because you can do more and better work with more colleagues and resources, or whatever. In physics, the answer seems to be no: on average, moving from a low-ranking university to an elite one had no effect on citations. Moving in the opposite direction had a slight negative effect. Confounding variables loom large here, of course. People sometimes move institutions for reasons that are correlated with other factors affecting their scientific impact, so these data can’t isolate any causal effects of institutional affiliation. Further, the dataset included relatively few moves between institutions of very different ranks, which means it’s not ideal even for estimating correlations between changes of affiliation and changes in impact. But FWIW, the findings jive with my own anecdotal experience: people vary in their productivity and impact (however measured), for all sorts of reasons, but are mostly about as productive at their old institution as at their new one.
A clever way of using phylogenetic diversity metrics to quantify the interdisciplinarity of grant proposals. Finds that more interdisciplinary work is significantly less likely to be funded, even after controlling for some obvious confounding variables. But the effect is small (at least to my eyes): expected funding probability drops from 0.24 for the least-interdisciplinary proposals to 0.18 for the most-interdisciplinary proposals.
An interview with Shahid Naeem, focusing on the backstory of his pioneering biodiversity-ecosystem function experiment in the Ecotron (Naeem et al. 1994 Nature). Fun story! I was surprised where the initial germ of the idea for the experiment came from. And I wasn’t aware that Shahid and the other postdocs–Sharon Lawler and Lindsey Thompson–had to talk John Lawton into dong it. It’s hard to believe now, when so many people have made their careers studying BEF, but John worried that the postdocs would be risking their careers by doing this experiment instead of a safe topical global warming experiment. Often what’s best for your career is to do something really new and controversial rather than something safe. Also, if you need reassurance that it’s ok to submit a conference abstract with only preliminary results, read this interview. 🙂 Finally, I love this line, especially since it comes from a guy who considered himself a field ecologist at the time:
Working in growth chambers is much more difficult than working in the field.
arXiv is considering a fundraising effort to update the service. I was interested to see the results of their user survey. Glad to see only a minority of users want arXiv to build an article commenting system. Extensive experience shows that basically nobody would use a commenting system and the vast majority of preprints would get no substantive comments. As for the desire of many users for filtering services that will identify the best preprints, others are already doing that so why shouldn’t arXiv just leave it to them?
The Dutch government will spend 8M euros on a national anonymized survey of Dutch scholars on their research practices, and on direct replications of important “cornerstone” research that has been relied on for policymaking or attracted significant media attention. (ht Marginal Revolution)