Some asymptotic and large deviation results in Poisson approximation (Q1203672): Difference between revisions

Let \(X_{n1},\dots,X_{nn}\), \(n\geq 1\), be independent random variables with \(P(X_{ni}=1)=1-P(X_{ni}=0)=p_{ni}\) such that \(\max\{p_{ni}\): \(1\leq i\leq n\}\to 0\) as \(n\to\infty\). Let \(W_ n=\sum_{1\leq k\leq n} X_{nk}\) and \(Z_ \mu\) be a Poisson random variable with mean \(\mu\). The paper deals with the asymptotic behavior of (i) \(r_ n=Eh(W_ n)-Eh(Z_ \lambda)\) and (ii) \(d_ n=\sum_{k\geq 0} h(k)| P(W_ n=k)-P(Z_ \lambda=k)|\) as \(n\to\infty\), where \(\lambda=EW_ n\) and \(h(k)\geq 0\), \(k\geq 0\). \(d_ n\) is the variation distance between the distributions \({\mathcal L}(W_ n)\) and \({\mathcal L}(Z_ \lambda)\). Theorem. Let \(E| h(Z_ \gamma+4)|<\infty\) for some \(\gamma>0\) and \(\lambda\to0\) as \(n\to\infty\). Then (i) \(r_ n\sim-2^{-1}\bigl(\sum_{i=1}^ n p_{ni}^ 2\bigr)(h(2)- 2h(1)+h(0))\) as \(n\to\infty\), where \(h(2)-2h(1)+h(0)\neq 0\); (ii) \(d_ n\sim 2^{-1}\bigl(\sum_{i=1}^ n p_{ni}^ 2\bigr)(h(2)+2h(1)+h(0))\) as \(n\to\infty\), where \(h(0)+h(1)+h(2)>0\). The authors also consider the case \(\lambda\to a\) as \(n\to\infty\), where \(a>0\). It is also proved that \[ {{P(W_ n>x)} \over {P(Z_ \lambda>x)}}-1\sim {{-x^ 2} \over{2\lambda^ 2}} \sum_{i=1}^ n p_{ni}^ 2 \qquad\text{as } n\to\infty, \] if \(\lambda\) remains bounded and \(x=o\bigl(\lambda\bigl(\sum_{i=1}^ n p_{ni}^ 2\bigr)^{- 1/2}\bigr)\) as \(n\to\infty\).