Rates of convergence of means of Euclidean functionals (Q2471129): Difference between revisions

Let \(X_1,\dots, X_n\) be i.i.d. random vectors with values in \(\mathbb{R}^d\). A special class of functions \(L(X_1,\dots, X_n)\), called Euclidean functionals, have been studied in a series of papers of a number mathematicians. In the paper the rate of convergence of mathematical expectations \(EL(X_1,\dots X_n)/b_n\), \(n\to\infty\), is investigated. Constants \(b_n\), \(n= 1,2,\dots\) are of a special form, for example \(n^{(d-p)/d}\) with \(p\in [1,d)\). Proofs are based on ideas from \textit{C. Redmond} and \textit{J. E. Yukich} [Ann. Appl. Probab. 4, No. 4, 1057--1073 (1994; Zbl 0812.60033) and Stochastic Processes Appl. 61, No. 2, 289--304 (1996; Zbl 0849.60030)]. All results are rather bulky to be cited here. The sum of the \(p\)th power-weighted length of the edges in a minimal tour \(\pi\) is an Euclidean function. The minimal tour \(\pi\) is a permutation on \(\{1,\dots, n\}\) with \(\pi(n+ 1)= \pi(1)\) such that \[ \sum^n_{k=1} |X_{\pi(k+1)}- X_{\pi(k)}|^p= \min\Biggl\{\sum^n_{k=1} |X_{\pi'(k+1)}- X_{\pi'(k)}|^p:\pi'\text{ a permutation on }\{1,\dots, n\}\Biggr\}. \]