SIMPSON'S PARADOX: CONFOUNDING AT ITS MOST CONFUSING?
Confounding is at the core of practical statistics. Whenever one is confronted with data, the manifest key aim is to locate any 'confounders' or 'lurking variables' (1).
That is, the same as a detective in a crime story, the statistician is hunting - not always easily - for hidden causes, concealed connections.
This Endgames problem refers to stratification as seeming to control potential confounders. Simpson's Paradox (2) is all about the dilemma of confounding; I find it hard to understand (it is terribly counter-intuitive).
There ought to be an endgame on Simpson's Paradox, just because it is so absolutely treacherous!
(1) A Clinician's Guide to Statistics. Epidemiology in Mental Health. S. Nassir Ghaemi. Cambridge 2009. Chapter 2.
(2) Statistical methods in medical research. 4th Edition. 2002. Blackwell. pg. 517.
Competing interests: No competing interests