The Matrix Sequential Probability Ratio Test and Multivariate Ruin Theory
摘要
The matrix sequential probability ratio test (MSPRT) is a statistical method to decide which law governs a collection of independent and identically distributed data amongst a finite set of possibilities. By focusing on the case where the possible laws are exponentially tilted versions of each other, in this note we exhibit novel links between the MSPRT and multivariate risk processes with common shocks, as well as with one-dimensional renewal theory.