Skip to main content

Sample Size and Sample Bias

This video presents the same text shown beside it, spoken and on screen. It adds nothing the text does not say.

State

Samples fail small — too few cases — or crooked: systematically unrepresentative. Bias is worse; no sample size cures it.

Show

The 1936 Literary Digest poll: 2.4 million responses, a skewed sampling frame, and heavy nonresponse bias. Landslide predicted; landslide happened — the other way.

Watch for

Big and biased beats small and biased only at being confidently wrong.