A distributed M-ary hypothesis testing problem with correlated observations

The distributed M-ary hypothesis testing problem of detecting one of M₀ events by an (N+1)-person hierarchical team, when the observations are correlated, is examined. In this problem, each of the N subordinate decision makers (DMs) transmits one of a prespecified set of messages based on their data to a primary decision maker who, in turn, combines the messages with his or her own data to make the final team decision. The necessary conditions for the optimal decision rules of the DMs are derived. A nonlinear Gauss-Seidel iterative algorithm is developed for the person-by-person optimal decision rules, and its monotonic convergence to the person-by-person optimum is established. A fast approximation algorithm is proposed for computing certain conditional probabilities arising in the person-by-person optimal decision rules. The algorithms are illustrated with several examples, and implications for distributed organizational designs are pointed out