scientific article; zbMATH DE number 1069536

From MaRDI portal
Publication:4356494





Given independent random variables \(X_1,\dots,X_n\) with the distribution \(\text{Pr} (\{X_i=1\})=1-\text{Pr} (\{X_i=0\})=p_i\) for \(i=1,\dots,n\), let \(a:=\sum_{i=1}^np_i\), \(\sigma^2:=\sum_{i=1}^np_i(1-p_i)\), \(\gamma:= \sigma^{-3}\sum_{i=1}^np_i(1-p_i)(1-2p_i)\) and \(S:=\sigma^{-1} \sum_{i=1}^n(X_i-p_i)\). Also, let \(x_m:=\sigma^{-1}(m-a+0.5)\) for \(m=0,1,\dots\) and let \(u(x):=(1+9\gamma^{-2}+6x\gamma^{-1})^{1/2}- 3\gamma^{-1}\) for \(x\geq -\gamma/6-3/(2\gamma)\). It is proved that \(|\text{Pr} (\{S\leq m\}) -\Phi (u(x_m))|<0.03\gamma^2+0.25\sigma^{-2}+0.75\sigma^{-3}\) for \(m=0,1,\dots\) provided \(\sigma\geq 10\), \(\gamma >0\) and \(x_m\geq \gamma/6 -\sqrt{2}\), where \(\Phi\) is the distribution function of a standard normal random variable. For the entire collection see [Zbl 0870.00038].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4356494)