No label defined (Q3683335)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3905636
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 3905636 |
Statements
1984
0 references
selection procedures
0 references
control value
0 references
two-stage sampling procedure
0 references
largest parameter
0 references
p-value
0 references
Operating characteristics
0 references
monotone likelihood ratio property
0 references
uniformly most powerful tests
0 references
[For the entire collection see Zbl 0542.00012.] Let \(\pi_ 1,...,\pi_ K\) be K populations associated with unknown parameters \(\theta_ 1,...,\theta_ K\in \Omega \subset R\) and let \(\theta_ 0\in \Omega\) be a known control value associated with a population \(\pi_ 0\), say. A population \(\pi_ i\) is considered better than \(\pi_ 0\) if \(\theta_ i>\theta_ 0.\) In this paper a two-stage sampling procedure is considered to find whether there is a population better than \(\pi_ 0\) and in the affirmative, to select the population associated with the largest parameter. Let \(x_ 1,...,x_ K\) and \(y_ 1,...,y_ K\) be samples drawn in the two stages from \(\pi_ 1,...,\pi_ K\), respectively. Let \(H_ i\) denote the hypothesis that \(\theta_ i\leq \theta_ 0\) and \(K_ i\) denote the alternative hypothesis that \(\theta_ i>\theta_ 0\). Let \(\phi_ i=\phi_ i(x_ i)\) and \(\psi_ i=\psi_ i(x_ i,y_ i)\) be randomized or non-randomized tests for \(H_ i\) versus \(K_ i\) and let \(p_{\phi_ i}(x_ i)\) be the p-value of \(x_ i\) under \(\phi_ i\) and \(p_{\psi_ i}(x_ i,y_ i)\) be the p-value of \((x_ i,y_ i)\) under \(\psi_ i.\) A procedure D is given as follows: At stage 1, select \(\pi_ i\) if the p-value of \(x_ i\) is \(\leq \alpha_ 0\), a pre-determined number. If none (exactly one) of the populations is selected, stop and decide that none is better than \(\pi_ 0\) (this one is better than \(\pi_ 0\) and is the best population). Otherwise proceed to stage 2 and select the population associated with the smallest p-value \(p_{\psi_ i}(x_ i,y_ i).\) Operating characteristics of the procedure D are given. It is shown that under the assumption that the underlying distribution of \((x_ i,y_ i)\) has monotone likelihood ratio property in \(\theta_ i\) every procedure (\(\phi\),\(\psi)\) in D is dominated by \((\phi,\psi^*)\), where \(\psi^*\) consists of uniformly most powerful tests, if \((x_ i,y_ i)\) are based on the same total number of observations for \(i=1,...,k\). If \(\theta_ 0\) is unknown same obvious changes are made in the procedure D.
0 references