Negative association, ordering and convergence of resampling methods (Q2313285)

Let \(\mathcal{X}\) be an open subset of \(\mathbb{R}^d\), \(\mathbb{X}\) be its Borel \(\sigma\)-algebra, \(\mathcal{P}(\mathcal{X})\) be the set of probability measures on the measurable space \((\mathcal{X},\mathbb{X})\), and let \(\mathcal{P}_f(\mathcal{X})\subset\mathcal{P}(\mathcal{X})\) be the subset of measures in \(\mathcal{P}(\mathcal{X})\) whose support is a finite set. In addition, let \[ \mathcal{Z}^N=\Big\{(x,w)\in\mathcal{X}^N\times \mathbb{R}_+^N:\sum_{n=1}^N w_{n}=1\Big\},\ \mathcal{Z}=\bigcup_{N=1}^\infty\mathcal{Z}^N. \] According to the authors, a resampling scheme is a mapping \(\varrho:[0,1]^\mathbb{N}\times\mathcal{Z}\rightarrow\mathcal{P}_f(\mathcal{X})\) such that, for any \(N\geqslant 1\) and \(z=(x,w)\in \mathcal{Z}^N\), \[ \varrho(u,z)=\frac{1}{N}\sum_{n=1}^N\delta\left(x_{a_N^n(u,z)}\right), \] where for each \(n\), \(a_N^n:[0,1]^\mathbb{N}\times\mathcal{Z}^N\rightarrow\{1,2,\dots,N\}\) is a measurable function, and \(\delta(x)\) denotes the Dirac measure at point \(x\). The authors of the paper study the convergence and the convergence rates for resampling schemes. In the paper, the authors present the main results obtained. The detailed proofs of the presented results can be found in the \textit{Supplementary Material} (see \url{arXiv:1707.01845}).