How to Lie with Statistics

How to Lie with Statistics — to Yourself and Others

  • Drawing conclusions from statistics is tricky. It’s easy for people to lie with statistics — to others and themselves.
  • People set forth bad arguments because
    • they’ve made honest mistakes in their reasoning
    • they’re biased
    • they’re deliberately trying to mislead people
  • Assume the following percentages for a country’s regions:
  • The percentages are correlated: the larger the percent vaccinated, the larger percent who got Covid. Indeed, the correlation coefficient equals +1, a “perfect correlation.” (Possible values run from -1 to +1.)
    • Correlation[{60, 55, 50, 45, 40}, {40, 30, 20, 10, 0}] = 1
  • It’s natural to infer that getting vaccinated makes it more likely that people get Covid or, at least, that vaccinations provide no protection.
  • But there’s a problem.