Weekly Links, July 15 -- U.S. Edition: what does police bias have to do with colliders, questions on POTUS publishing in JAMA.
What’s JAMA’s new impact factor now that POTUS has published a paper there? As you probably heard, Mr. Obama published a paper in the Journal of the American Medical Association this week, describing the progress to date of the US Health Care Reform and outlining the next steps. I have so many questions: was the review process (if there was one) double blind? Was he first rejected from NEJM? Was there a revise and resubmit? Was Obama totally nice to that rude referee #2, so that his paper could get published without further hassle? If you’re a handling editor or a referee, we want to hear from you (anonymously or not)...
An NBER paper by Roland G. Fryer, Jr. on gun violence in the US by race has made waves, mainly because people have hard time believing the finding that blacks are less likely to be fatally shot during police stops/arrests than whites. Uri Simonsohn on datacolada.org has expressed disbelief and pointed to the noise in the data to state that they are consistent with “Blacks being 50% less likely or 80% more likely to be shot than whites.” He, and others, such as Andrew Gelman, suggested selection bias as the main potential problem: in the lingo of causal inference using directed acyclic graphs (or DAGs), this sounds like a collider problem. Suppose that acting smarts and looks are completely uncorrelated with each other. However, each of them affect your chances of being a Hollywood star. If you condition your analysis on being a Hollywood star, you can generate a spurious negative correlation between beauty and acting ability: the movie star variable is a collider, conditioning on which is a no-no. So, Blacks may be more likely to be shot, but once you condition on “shoot or don’t shoot “ situations, then you might find that they’re equally or even less likely to be shot…I don’t know the answer at all, but Fryer answered some questions here…
The Development Impact Evaluation group (DIME) is now hiring field coordinators in Rwanda, Tanzania and Nepal.