A/B testing — the p-value that a peeked experiment inflates
A/B tests promise a clean causal read from randomization, but stopping early when the result looks good — peeking — silently multiplies the false-positive rate the test claims to control.
Not yet scoped.
Why
The PoC replays a null experiment with daily peeking and shows how often it 'wins', then contrasts fixed-horizon and sequential designs that actually hold the error rate.
How it works
Not yet built.
Where it lands in Jayverse
Rabbit/Verex: fix the test horizon before running any web-app A/B test. Use a fixed sample size or an explicit sequential design — no stopping early because a metric looks good, since peeking silently inflates the false-positive rate.
gitboard: note which design backed each reported experiment win. Fixed-horizon vs sequential, since a peeked result changes the true false-positive rate behind any number gitboard displays.
Key expressions
Words and phrases from this page worth keeping, with the Korean meaning and the sentence they come from.
Expression
뜻 · 쓰이는 자리
peeking
중간에 몰래 들여다보기 · 실험 결과를 정해진 종료 전에 미리 확인하는 행위. "stopping early when the result looks good"
silently multiply
소리 없이(티 안 나게) 배가시키다 · 오류율이 알아채지 못하는 사이 커질 때. "silently multiplies the false-positive rate"
a clean causal read
깔끔한 인과적 해석 · 무작위 배정 덕에 원인과 결과를 명확히 볼 수 있을 때. "a clean causal read from randomization"
replay
(같은 상황을) 재현하다 · 실험을 다시 시뮬레이션해볼 때. "replays a null experiment with daily peeking"
hold the error rate
오류율을 (목표 수준으로) 유지하다 · 통계적 방법이 원래 약속한 오류 수준을 지킬 때. "designs that actually hold the error rate"