Parameters
%
%
one-tail
20% = 80% power
How it works: SPRT tests hypotheses sequentially, allowing you to stop early when evidence is sufficient. Continue collecting data until you hit an upper boundary (reject null) or lower boundary (accept null).
Results
Upper Boundary (LR)
—
Stop & reject null (significant)
Lower Boundary (LR)
—
Stop & accept null (not significant)
Expected Sample Size
—
Expected observations per variant
Savings vs Fixed-Horizon
—
Reduction in required sample size