Abstract
The paper “An estimate of the science-wise false discovery rate (SWFDR) and application to the top medical literature” by Jager and Leek provides an interesting perspective on FDRs in published medical research studies. As the authors point out, the distribution of p-values under the null hypothesis in these studies may well depart from uniform, and for this reason the authors explore the consequences of “p-hacking” and rounding of p-values on the FDR. Even after accounting for p-hacking, the authors conclude that the FDR is not alarmingly high. However, the authors appear to ignore other possible sources of departure from uniformity such as bias, model misspecification, and measurement error.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have