Optimality of the sequential probability ratio test for nonstationary observations

Bayesian analysis is used to show that Wald´s sequential probability ratio test with varying thresholds is optimal for the nonstationary situation, where the observed samples are independent but not identically distributed. Some important properties useful for the design of the test thresholds are discussed. Wald´s lower bound, generalized to the nonstationary situation, is also presented. The results have important applications in situations where the observed signal is time-varying. such as in radar signal processing, image processing, and spread spectrum communications