Evan Miller

1 note in the collection. Writes at evanmiller.org (opens in a new tab)

How Not To Run an A/B Test Oct 12 7 min
Miller argues that peeking at A/B test results and stopping once they look significant inflates false positives, so sample size must be fixed in advance.
Related
- Taste for MakersPaul Graham
- Take Something AwayTed Lamade
- Notes on Incident ResponseSlava Akhmechet
- Measurable does not mean importantFay
Read the original of How Not To Run an A/B Test (opens in a new tab)