Order Sampling Design with Prescribed Inclusion Probabilities

From MaRDI portal





This paper deals with order sampling with a fixed distribution shape. Let us consider a sequence \(Q_{1},Q_{2},\ldots\) of independent positive random variables with distribution functions \(F_{1},F_{2},\ldots\). Let the inclusion probability of unit \(i\) be a function of the \(\theta\) parameters: \[ g_{i}(\theta_{1},\ldots,\theta_{N})=\int_{0}^{\infty}(1-F_{n}^{N- 1}(\{\theta_{i}\}^{c},t))f_{i}(t)dt, \] where \(\{\theta_{i}\}^{c}\) denotes \((\theta_{1},\ldots,\theta_{i-1},\theta_{i+1},\ldots, \theta_{N})\), and thus \( F_{n}^{N- 1}(\{\theta_{i}\}^{c},\cdot)\) is the distribution of the \(n\)-th order statistic of \(Q_{1},\ldots, Q_{i- 1}, Q_{i+1},\ldots, Q_{N}\), and \(f_{i}(t)\) is the density of the distribution function \(F_{i}(t)\). The authors prove that for any set of values of \(z_{1},\ldots,z_{N}\) such that \(0<z_{i}<1\) and \(\sum_{i=1}^{N}z_{i}=n\), the system of equations \(g_{i}(\theta_{1},\ldots,\theta_{N})=z_{i}\), \(i=1,\ldots,N\), has a unique solution in \(\theta_{1},\ldots,\theta_{N-1}\) for any fixed \(\theta_{N}\in(0,\infty)\). Furthermore, it is always possible to choose \(\theta_{N}\) such that this solution satisfies \(\sum_{i=1}^{N}F_{i}(1)=\sum_{i=1}^{N}H(\theta_{i})=n\), where \(H(t)\) is a fixed distribution shape, and \(F_{i}(t)=H(t\cdot \theta_{i})\), \(i=1,\ldots,N\).











This page was built for publication: Order Sampling Design with Prescribed Inclusion Probabilities

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4416171)