Approximation of a sample by a Poisson point process (Q2567785): Difference between revisions

The distribution of a sum of \(n\) independent random variables \(X_i\) has been approximated by that of the `accompanying' infinitely divisible law, with an error bound uniformly over all distributions: see, for example, \textit{È. L. Presman} [Theory Probab. Appl. 18, 378--384 (1973); translation from Teor. Veroyatn. Primen. 18, 396--402 (1973; Zbl 0307.60030)]. The accompanying law is a compound Poisson distribution with intensity measure \(\sum^n_{i=1}{\mathcal L}(X_i)\) (possibly also with centering), so that the same theorem implies the approximation of \({\mathcal L}(\sum^n_{i=1} g(X_i))\) by a compound Poisson distribution with intensity measure \(\sum^n_{i=1}{\mathcal L}(g(X_i))\), for any \(g\). This can in turn be viewed as a measure of the closeness of the distribution of the sample with values \(X_1,\dots, X_n\) to that of the Poisson process with intensity measure \(\sum^n_{i=1}{\mathcal L}(X_i)\).