Also this week: being black at Michigan, handy new R package, subtle overfitting, the myth of the myth of the hot hand (as Andrew Gelman likes to say), the ROC curves of science, the birds and the bees vs. “the birds and the bees”, pun-ishing bad statistics, and more. Oh, and Meg vs. Jeremy.
This incredibly powerful video features 16 black students who share what their experiences being black in America and at the University of Michigan. It covers their experience in their lives out of the classroom, but also in the classroom. It’s long (20 minutes) but very much worth watching.
This new R package, cowplot, by Claus Wilke looks like it will help avoid the need to move to Illustrator, Powerpoint, etc. to combine multiple ggplot2 plots into one and add A, B, C, D, etc. labels to the plot. This should make it easier to create multipanel figures all in R!
Ed Yong has a write up of a new PNAS paper that asks how African grasslands support so many herbivores. It’s a neat story about using molecular genetics to uncover an impressive degree of niche partitioning. The PNAS paper is by Kartzinel et al.
If you use cross-validation as a model selection tool, the odds are that you’re subtly overfitting your data. Put another way, the distinction between model selection and model fitting is fuzzy. Interesting, I’d never thought of this. (ht Andrew Gelman)
Speaking of statistics, here are ecological statistician Mark Brewer’s top 10 tips for reviewing the statistics in ecology papers. I especially like numbers 2, 3, 7, and 8.
Whatever happened to the population bomb? Newspaper article plus video, could be good fodder for a class discussion. Related: my review of a popular history of the famous Ehrlich-Simon bet on the population bomb. (ht Emily Weigel, via Twitter)
Stephen Heard elaborates on his argument for spreading grant funding fairly evenly within scientific fields, considering the effects of various complications. Nice example of using a simple toy model (really, a series of increasingly-elaborate toy models) to think through a problem.
Speaking of grant funding: Mike the Mad Biologist with interesting thoughts on whether and why the Human Genome Project was worth it, and on Big Science more generally.
This is old but it’s very good and I missed it at the time: the ROC curves of science. Or, why all those calls to deal with the “replication crisis” by making statistical inference more conservative are misguided (or at least come with a substantial downside).
A good news article on statistical evidence for hot streaks in sports. Could be good fodder for undergrad stats courses.
An interview with paleontologist Jack Horner on his experiences as a science consultant on the Jurassic Park films. Resonates with a book I read (and that I keep meaning to review for the blog) about science in the movies.
Tough times at the University of Wisconsin these days.
Scientific papers aren’t as weird as they used to be. Although my question is which of these papers, if any, was seen as weird at the time.
Collect ALL the stats puns! I like “Bayes Watch” and “Hawaii 0.05”.🙂
Hoisted from the archives: