The succession procedure for interim analysis: extensions for early acceptance of Ho and for flexible times of analysis.

The succession procedure is a new method for group sequential analysis based on a succession of significant tests. In this paper, the procedure is extended by the addition of a lower boundary, which allows the early acceptance of the null hypothesis Ho according to the stochastic curtailing approach. The succession procedure is also adapted to the type I error spending rate function alpha *(t), leading to the method of semisequential analysis for which the times of analysis need not be specified in advance.