Ben Goldacre at the Guardian on a new paper in Nature Neuroscience that shows that academic psychologists publishing in leading neuroscience journals do not correctly compute the standard errors when doing "difference-in-differences", which is to say, when comparing changes in some outcome in response to treatment for two different groups.
Economics has many problems, but thankfully this is not one of them. Perhaps we are saved here by the fact that we shoehorn all empirical analyses into the context of linear regression, where the difference-in-differences estimand corresponds to the coefficient on an interaction term and the regression package automatically deals with the standard error issues that apparently bedevil the psychologists.
Hat tip: Julia Lane
Who was my favorite student this term?
4 months ago