A limit theorem for continuous selectors (Q312333): Difference between revisions

A continuous selector is a continuous function \(H:\mathbb{R}^n\mapsto\mathbb{R}\), where \(H(x_1,\ldots,x_n)\in\{x_1,\ldots,x_n\}\) for each \((x_1,\ldots,x_n)\in\mathbb{R}^n\). For each such function, an operator \(\mathbf{H}\) acting on random variables \(X\) is defined by \(\mathbf{H}(X)=H(X_1,\ldots,X_n)\), where \(X_1,\ldots,X_n\) are independent copies of \(X\). The main result of the present paper is a limit theorem for the iterates \(\mathbf{H}^{(N)}(X)\) as \(N\rightarrow\infty\). The authors show that for each non-degenerate continuous selector \(H\) there is a unique \(\omega_H\in[0,1]\) such that \(\mathbf{H}^{(N)}(X)\) converges in distribution to the quantile of \(X\) corresponding to \(\omega_H\). In proving this, the authors use the concept of Sperner statistics. A Sperner family in \(\{1,\ldots,n\}\) is a collection of subsets \(A_1,\ldots,A_k\) of \(\{1,\ldots,n\}\) such that no \(A_i\) is contained in any other \(A_j\). Each such Sperner family is associated with the Sperner statistic \(H_\mathcal{S}\) in \(\mathbb{R}^n\) given by \(H_\mathcal{S}=\max(\min_{A_1},\ldots,\min_{A_k})\). The authors show that any continuous selector is a Sperner statistic, and that the converse statement also holds. This characterization is used in the proof of the limit theorem discussed above. The motivation for this problem is provided by a discussion of a randomized version of Zermelo's algorithm.