scientific article; zbMATH DE number 3942778
[For the entire collection see Zbl 0564.00019.] Let \(\pi_ 1,...,\pi_ k\) denote k populations with respective continuous distribution functions \(F_ 1(x),...,F_ k(x)\). Let a number \(\alpha\) \((0<\alpha <1)\) be given. The problem considered is that of selecting the population associated with the largest \(\alpha\)-quantile and of simultaneously giving an interval estimate for it. Let \(x_{\alpha}(F_ i)\) denote the \(\alpha\)-quantile of \(F_ i(x)\), \(i=1,...,k\). Write \(F_{[i]}\leq F_{[j]}\) if \(x_{\alpha}(F_{[i]})\leq x_{\alpha}(F_{[j]})\) where \(x_{\alpha}(F_{[1]})\leq...\leq x_{\alpha}(F_{[k]})\) are the k \(\alpha\)-quantiles in numerical order. Let CS denote the event of correctly selecting the population corresponding to \(F_{[k]}\) and CI the event \(x_{\alpha}(F_{[k]})\in I\), where I is the proposed confidence interval. One formulation is as follows: Let \(\epsilon_ 1\) and \(\epsilon_ 2\) be given numbers such that \(\epsilon_ 1\leq \alpha \leq 1-\epsilon_ 2\). Let \(\underline x_{\beta}(F_ i)=\inf \{x:\quad F_ i(x)=\beta \},\) \(\bar x(F_ i)=\sup \{x:\quad F_ i(x)=\beta \},\) and \(\underline F=\underline F(x)=\min (F_{[1]}(x),...,F_{[k-1]}(x))\) for all real x. Define the closed interval J by \(J=[\bar x_{\alpha -\epsilon_ 1}(F_{[k]}),\quad \underline x_{\alpha +\epsilon_ 2}(F_{[k]})].\) Let \(d=\inf_{x\in J}(F(x)-F_{[k]}(x)).\) A procedure R is proposed such that \(P(CS\cap CI)\geq P^*\) for all \(F_ 1,...,F_ k\) with \(d\geq d^*\), where \(P^*\) \((1/k<P^*<1)\) and \(d^*\) are preassigned given numbers. Three other formulations are also considered. A modified method of applying the procedure based on n independent observations from each of the k populations is given and is illustrated with an example. Tables are provided to implement the procedure when \(\alpha =0.5\).
- Nonparametric Selection of the Smallest-Dispersion Population
- scientific article; zbMATH DE number 218781
- Selection procedures for scale parameters when the populations have a common known quantile
- A new procedure for selecting good populations
- Rizvi and Sobel Type Nonparametric Selection Procedures: Review and Extensions
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 Q3713372)