Simpson's Paradox and the implications for medical trials

Norman Fenton, Martin Neil, Anthony Constantinou

This paper describes Simpson's paradox, and explains its serious implications for randomised control trials. In particular, we show that for any number of variables we can simulate…