Sequential testing for multinomial categories using Dirichlet Integrals.
by
Hokwon Cho
University of Nevada, Las Vegas
Coauthors: Hai Zhen

An inverse-type sequential sampling procedure is considered for testing k (<oo) multinomial categories with non-zero probabilities. An efficient stopping rule is devised to satisfy a pre-determined level of confidence, P*-condition. For developing the procedure, Dirichlet Type-II integrals will be primarily used with the decision theoretic point of view. Optimal sample sizes are obtained along with the desirable probability level PCD, the probability of correct decision. Various configurations on “wheel of fortune” will also be considered for an illustration. Finally, a comparison with the existing method and the developed procedure will be discussed via Monte Carlo experimentation.

Date received: March 30, 2007