A Simple, General, and Efficient Method for Sequential Hypothesis Testing: The Independent Segments Procedure

We propose a new sequential hypothesis testing procedure in which data are collected and analyzed in a series of independent segments. As in fixed-sample hypothesis testing and in previous sequential procedures, the overall alpha level can be set to any desired value. Like other sequential procedures, the independent segments procedure generally requires smaller samples than fixed-sample procedures-often approximately 30% smaller-to achieve the same alpha level and statistical power. Relative to other sequential procedures, the new method has the advantages that it is simpler to use, requires fewer assumptions, and can be used with a wider array of statistical tests. Thus, in some circumstances the independent segments procedure may provide an attractive option for increasing the efficiency of statistical testing. Translational Abstract We propose a new procedure for conducting scientific research in which data are collected and analyzed in a series of independent subexperiments or segments. Compared with the standard methods that are most commonly used, the new procedure can provide the same statistical power with smaller samplesoften approximately 10-30% smaller. Thus, the new procedure can help to increase the efficiency of scientific research.