On the robust discrimination of Poisson random counting measures (Corresp.)

Statistical decisions are considered between two hypotheses which consist of classes of Poisson distributions for random counting measures. Each Poisson distribution is generated by an intensity measure on a general observation space. The classes are specified by Choquet capacity bounds on the intensity measures. This problem was first posed and studied by Geraniotis and Poor. A minimax Neyman-Pearson result for error probability performance is the main contribution. Recursive computation of the minimax test statistic is also investigated.