2001 Volume 28 Issue 1 Pages 23-35
In the present paper, the ordinal Play-the-Winner sampling with inverse stopping rules is generalized in view of the number of treatments, the introduction of an available sample size and so on. The probability of the correct and wrong selection and the expected number of total sample size are formulated by using the negative binomial distribution without any use of the asymptotic theory. Monte Carlo experiments are performed for some practical situations to look into several properties numerically. As a criterion for optimality of such plans, the minimization of the expected loss proposed by Colton is used with several probabilities and expectation of random numbers to design an optimal Generalized Play-the-Winner plan.