scientific article; zbMATH DE number 3901855

From MaRDI portal
Publication:3680112





Conditional on the value of \(\theta >0\), \(x_ 1, x_ 2,..\). are i.i.d. real random variables having a gamma distribution with density function \(p_{\theta}(x)=(\theta \lambda)^{\lambda}(\Gamma (\lambda))^{- 1}x^{\lambda -1}\exp (-\theta \lambda x);\) \(x\geq 0\), \(\lambda >0\), \(\theta >0\). Sequential estimation procedures of the form (\({\hat \theta}{}_{\tau}, \tau)\) are considered. To measure loss due to estimation and the cost of sampling a sequence of functions, indexed by p, is used: \(g_ n(\theta,{\hat \theta}_ n)=(\theta -\theta_ n)^ 2/\theta^{-p}+c(n),\) \(p\geq 0\). The notion of monotone case for the posterior risk is used. When the prior distribution of the parameter \(\theta\) has the form \(p(\theta)=\beta^{\alpha}(\Gamma (\alpha))^{-1}\theta^{\alpha - 1}\exp (-\beta \theta),\) \(\alpha >0\), \(\beta >0\) and \(\alpha +\beta >p\geq 0\) then a Bayes sequential estimator is obtained with optimal stopping time when \(1\leq p\leq 2\) and quasi-optimal stopping time when \(0\leq p<1.\) The efficiency of the sequential procedure for \(p=0\) and \(c(n)=c\cdot n\) is investigated. The case \(\lambda =1\) also can be found in \textit{C. P. Shapiro} and \textit{R. L. Wardrop}, Ann. Stat. 8, 171-182 (1980; Zbl 0434.62062).











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 Q3680112)