35
Sampling Plan
•Perform analyses at sample sizes N1. . . NJ
–Can be randomly determined
•At each analysis choose stopping boundaries
–aj < bj < cj < dj 
•Compute test statistic T(X1. . . XNJ)
–Stop if T  < aj                     (extremely low)
–Stop if bj < T < cj        (approximate equivalence)
–Stop if T > dj                      (extremely high)
–Otherwise continue (with possible adaptive modification of analysis schedule, sample size, etc.)
•