ABSTRACT
Our discussion of testing hypotheses up to this point assumed a fixed sample size n and a critical boundary, either set arbitrarily or determined by specified error sizes α and β. In situations where data are collected sequentially, it may become clear early on what the situation is, whereas the test would call for proceeding to collect all the data before coming to any conclusion. If the evidence becomes overwhelming, one way or the other, before the entire sample of size n is observed, there is nothing to prevent one from stopping the sampling and coming to a conclusion; but if this is done, the planned error sizes are not applicable.
