MEPX
Chapter 9 of 10All chapters

Chapter 9 of 10

How statistics mislead

Usually without lying.

Common tricks

Truncated axes, cherry-picked start dates, percentages of different bases, and averages hiding a bimodal split. None of these require a false number.

  • Simpson's paradox: a trend in every subgroup can reverse when combined.
  • Testing many things guarantees some will look significant by chance.

Questions to ask

Who was measured, how were they chosen, what was the sample size, what is the uncertainty, and what is being compared with what.