Applications of two majorization inequalities to ranking and selection problems
Let \(\{ X_{i}\}_{i=1}^{\infty}\) be an infinite sequence of random variables. This sequence is said to be exchangeable if for all finite \(m\) and for all permutations \((\Pi_1,\ldots,\Pi_{m})\) of \((1,\ldots,m)\) the random vectors \((X_{\Pi_1},\ldots,X_{\Pi_{m}})\) and \((X_{1},\ldots,X_{m})\) have the same distribution. A random vector \({\mathbf X}=(X_1,\ldots,X_{m})\) is said to be exchangeable if there exists an exchangeable sequence \(\{ X_{i}^{\ast}\}_{i=1}^{\infty}\) such that \((X_1^{\ast},\ldots,X_{m}^{\ast})\) and \(X\) have the same distribution. The author proves the following statement: Let \({\mathbf X}=(X_1,\ldots,X_{m})\) be an exchangeable random vector. Then \[ P[X_{i} \in A,i=1,\ldots,m]\geq P[X_{i} \in A,i=1,\ldots,m_1]\cdot P[X_{i} \in A,i=1,\ldots,m-m_1] \] \[ \text{and}\qquad P[X_{i} \in A,i=1,\ldots,m]\geq P\{[X_{i} \in A,i=1,\ldots,m_1]\}^{m/m_1} \] hold for all Borel-measurable sets \(A\) and all \(m_1 <m.\) These inequalities are used in ranking and selection problems. The author found the upper and the lower bounds for the true probability function of the correct selection. He applied these results to the selection procedures under Bechhofer's indifference-zone formulation and Gupta's maximum type subset formula. For the entire collection see [Zbl 0990.62506].
This page was built for publication: Applications of two majorization inequalities to ranking and selection problems
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3843179)
