MEPX
Chapter 5 of 10All chapters
  1. 1What statistics is for
  2. 2Averages
  3. 3Spread
  4. 4Distributions
  5. 5Sampling
  6. 6Confidence and error
  7. 7Testing a claim
  8. 8Relationships between variables
  9. 9How statistics mislead
  10. 10Doing it well

Chapter 5 of 10

Sampling

Where conclusions are won or lost.

Randomness matters more than size

A random sample represents the population; a convenient one represents whoever was convenient. A huge biased sample is confidently wrong.

  • The 1936 Literary Digest poll had millions of responses and got the election backwards.
  • Non-response is a bias: the people who ignored you may differ systematically.

Sample size

Precision improves with the square root of the sample size, so quadrupling the sample halves the margin of error. Beyond a few thousand, extra respondents buy very little.