Joint distributions of runs in a sequence of multi-state trials (Q1962598): Difference between revisions

In recent years, exact discrete distribution theory of success runs has been developed. The runs most frequently used are: \(E_{n,k}\), the number of success runs of length exactly \(k\) in \(n\) trials; \(N_{n,k}\) (and \(M_{n,k}\)), the number of nonoverlapping (and overlapping) consecutive \(k\) successes in \(n\) trials; and \(G_{n,k}\), the number of success runs of length greater than or equal to \(k\) in \(n\) trials. \textit{J. C. Fu} and \textit{M. V. Koutras} [J. Am. Stat. Assoc. 89, No. 427, 1050-1058 (1994; Zbl 0806.60011)] considered a finite Markov chain imbedding technique to study the distributions of all the above statistics, and \textit{M. V. Koutras} and \textit{V. A. Alexandrou} [Ann. Inst. Stat. Math. 47, No. 4, 743-766 (1995; Zbl 0848.60021)] introduced a Markov chain imbeddable variable of binomial type (M.V.B.) to study \(N_{n,k}\), \(M_{n,k}\) and \(G_{n,k}\). In the present paper the authors extend the concept of M.V.B. They introduce a Markov chain imbeddable vector of multinomial type to study joint distributions of the above modes of runs in a sequence of multi-state trials. The evaluation of the joint distributions is performed recursively. Furthermore, they introduce the Markov chain imbeddable variable of returnable type, which enables the evaluation of the distribution of \(E_{n,k}\) in a sequence of Bernoulli trials.