Also this week: the history of #icanhazpdf, microbiome pet peeves, new evidence of widepread p-hacking, the difference between terrestrial and marine ecologists, using software to solve the wrong problem, and more. Oh, and never mess with a manatee.
Jacquelyn Gill has a post that provides an important reminder about the importance of considering lab safety. Fortunately, no one in my lab has had an accident as serious as hers, but someone did cut her hand (while making chemostats) enough to require an immediate trip to the doctor. As a grad student, I had to drive a student to urgent care after she sliced her foot open on something while wading into a lake in a limnology class. I sliced my own foot on a zebra mussel in a different lake. And when I worked as a technician between college and grad school, the other tech had what initially appeared to be a very, very bad cut to her hand, but that fortunately ended up not being very serious. So, it’s clear to me that things can go wrong and, combine that with me being the child of a nurse and a fireman (safety first!), and you’d think my lab would be all about the safety training. But Jacquelyn’s post has me realizing that I probably haven’t thought about this enough. We have some basic safety measures (especially that people should head into the field with a buddy, that you need to get off the lake at the first sign of a thunderstorm, and that no one is allowed to go out in the boat unless they can swim); at Georgia Tech I made sure everyone in the lab knew the number to call if there was an emergency and had this taped to the lab phones. (It wasn’t 911 because that would get Atlanta police, whereas GT police would be able to respond faster.) But how would people respond in a situation like the one the Gill Lab was in? I’m not sure. Then again, what sort of training could we do that would prepare folks for the wide variety of (fortunately unlikely!) situations that could arise? Definitely lots to think about!
On twitter, people use the #icanhazpdf hashtag to ask for pdfs that they can’t get on their own (usually via institutional access). For people at institutions, this is a way to bypass the InterLibrary Loan (ILL) system (and I think often results in getting a pdf more quickly). This paper has an interesting summary of #icanhazpdf, including its history and information on what is being requested. It also includes this depressing sentence:
The current scholarly publishing system is so broken that some researchers are forced to make requests like “Still looking for a pdf of my own paper! Please help.”
Sociologist Andrew Lindner on how a paper of his was scooped by a blog post, and what this says about the scholarly publishing system. I wouldn’t overgeneralize from what seems like an unusual coincidence, but still, interesting to think about. (ht Brad DeLong)
Noah Fierer’s pet peeves of microbiome studies.
The potted history of P-values, at least when told by certain sorts of Bayesians, is that they were an invention of R. A. Fisher that set scientific inference on the wrong basis for the better part of a century. For instance, Nate Silver spends a whole chapter on this potted history in his recent book. Statistician Stephen Senn corrects the historical record (emphasis in original):
Fisher did not persuade scientists to calculate P-values rather than Bayesian posterior probabilities; he persuaded them that the probabilities that they were already calculating and interpreting as posterior probabilities relied for this interpretation on a doubtful assumption. He proposed to replace this interpretation with one that did not rely on the assumption.
The upshot, Senn argues, is that Bayesians don’t really have a problem with P-values. Rather, they have a problem with other Bayesians. And many contemporary complaints about the evils of P-values are misdiagnosing the root of the problem. Go read and then join the (already lengthy!) discussion in the comments.
Marine ecologists are organism/system focused, terrestrial ecologists are question-focused. At least, that’s one way to interpret the fact that marine papers name the study organism(s) much earlier in the introduction than do terrestrial papers (Menguia & Ojanguren 2015, open access). Casey terHorst comments. Meg, what do you think the results would be if you looked at freshwater ecology? (ht @hughes_lab)
Head et al. 2015 (open access) text mined all open access papers on PubMed and examined the distribution of P-values <0.05 to look for evidence of p-hacking. It’s a very careful study, more careful than most of casual text mining on this topic that I’ve linked to in the past (somewhat to my regret). Turns out p-hacking is widespread across scientific disciplines covered by PubMed. The results also suggest that scientists mostly are studying real effects rather than chasing noise. Head et al. also found evidence of p-hacking in meta-analyses of sexual selection in evolutionary biology, but not enough to dramatically alter the conclusions of the meta-analyses.
A while back we did a post on ecologists who are awesome at things besides ecology. In the same spirit, I give you Ravens offensive lineman John Urschel, who is a serious mathematician. (ht Marginal Revolution)
Straight from the horse’s mouth: the relationship between academic economics and economics blogging. (ht Marginal Revolution)
This week in Treating the Symptom Not the Disease: dude, if you need software to tell you if a scientific paper was computer generated, your journal has problems no software can fix.
And finally, never mess with a manatee. 🙂