Simpson’s Paradox: How statistics can mislead
What hospital to go to: A or B. Well, hospital A’s patients survive 90% of the time and hospital B’s survive 10% of time. Clearly, hospital A is the right choice. Well, not necessarily. Hospital B’s patients could be a much sicker group to begin with. Should you buy stock in company A or company B. Well, A has higher margin and faster growth, So clearly company A. Well, not necessarily. Company A also has too much debt and its growth is all from acquisitions. Company B’s growth is organic and its balance sheet is debt free. Similarly, is there gender discrimination in at college x. Well, say 60% of women who apply are admitted while 90% of men are. Clearly there is discrimination, right? Not necessarily. Perhaps the women are applying to more competitive departments. It is extraordinarily easy to come to the wrong conclusion based on incomplete data. Demagogues love twisting your emotions with data that sound compelling when critical, granular data is omitted.