On a direct method of analysis and synthesis of the SPRT

The behavior of the sequential probability ratio test (SPRT) can be described using four performance measures, i.e., the probabilities of both types of error \alpha_{0}, \alpha_{1} , and the average test lengths (or sample sizes) E_{0}, E_{1} for the given decision problem. Whereas Bayes optimality of the SPRT is well-known from the work of Wald and Wolfowitz, there is no method to evaluate the performance measures except for Wald's approximations. We present a method to evaluate the performance measures given the thresholds of the SPRT.