Today, I uploaded a sequence of three working papers to my website at https://andershuitfeldt.net/working-papers/

This is an ambitious project that aims to change fundamental things about how epidemiologists and statisticians think about choice of effect measure, effect modification and external validity. A link to an earlier version of this manuscript was posted to Less Wrong half a year ago, the manuscript has since been split into three parts and improved significantly. This work was also presented in poster form at EA Global last month.

I want to give a heads up before you follow the link above: Compared to most methodology papers, the mathematics in these manuscripts is definitely unsophisticated, almost trivial. I do however believe that the arguments support the conclusions, and that those conclusions have important implications for applied statistics and epidemiology.

I would very much appreciate any feedback. I invoke "Crocker's Rules" (see http://sl4.org/crocker.html) for all communication regarding these papers. Briefly, this means that I ask you, as a favor, to please communicate any disagreement as bluntly and directly as possible, without regards to social conventions or to how such directness may affect my personal state of mind.

I have made a standing offer to give a bottle of Johnnie Walker Blue Label to anyone who finds a flaw in the argument that invalidates the paper, and a bottle of 10-year old Single Scotch Malt to anyone who finds a significant but fixable error; or makes a suggestion that substantially improves the manuscript.

If you prefer giving anonymous feedback, this can be done through the link http://www.admonymous.com/effectmeasurepaper .