Tests for means with additional information (Q3700618)

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 3926011
Language Label Description Also known as
default for all languages
No label defined
    English
    Tests for means with additional information
    scientific article; zbMATH DE number 3926011

      Statements

      Tests for means with additional information (English)
      0 references
      0 references
      0 references
      0 references
      1985
      0 references
      admissible Bayes tests
      0 references
      likelihood ratio test
      0 references
      Hotelling's T
      0 references
      square
      0 references
      tests for means
      0 references
      additional information
      0 references
      critical regions
      0 references
      Bayes tests
      0 references
      normal random samples
      0 references
      two step-down procedures
      0 references
      small sample powers
      0 references
      Assume \(Z=(X'Y')'\sim N_{p+q}(\mu,\Sigma)\). It sometimes appears that some observations on Y are missing but the corresponding observations on X are available. Some tests for means in this situation are discussed. Let \(\xi,\xi_ 1,...,\xi_ n\) be independent r.v.s, \(\xi \sim N_{p+q}(\delta,\Sigma)\) and \(\xi_ i\sim N_{p+q}(0,\Sigma)\), where \(\delta =(\delta '\!_ 1,\delta '\!_ 2)'\) and \(\Sigma =\left( \begin{matrix} \Sigma_{11}\\ \Sigma_{21}\end{matrix} \begin{matrix} \Sigma_{12}\\ \Sigma_{22}\end{matrix} \right)\), \(\delta_ 1:p\times 1\) and \(\Sigma_{11}:p\times p\). Also, let V, \(W_ 1,...,W_ m\) be independent r.v.s, \(V\sim N_ p(\delta_ 1,\Sigma)\) and \(W_ i\sim N_ p(0,\Sigma_{11})\), \(i=1,...,m.\) An analogous result to that of \textit{C. Stein} [Ann. Math. Stat. 27, 616- 623 (1956; Zbl 0073.143), see also Proc. 3rd Berkeley Sympos. math. Statist. Probability 1, 197-206 (1956; Zbl 0073.356)] is obtained as follows: for testing \(H_ 0:\delta =0\) vs \(H_ 1: \delta \neq 0\), the acceptance region of Hotelling's \(T^ 2\)-test based on \(\xi,\xi_ 1,...,\xi_ n\), which ignores the additional information provided by V, \(W_ 1,...,W_ m\), is given by \(\{(S,S^*,\xi,V):S\) is p.d. and \(\xi 'S^{-1}\xi \leq K\}\) where \(S=\xi \xi '+\sum_ i\xi_ i\xi '\!_ i\), \(S^*=VV'+\sum_ iW_ iW'\!_ i\), and this test is admissible. Some critical regions of admissible tests for determining unique Bayes tests in multivariate normal random samples are provided for the previous model. Several two step-down procedures are proposed. The results that these tests are admissible are obtained under certain conditions on \(\delta\) or \(\Sigma\). In addition, evaluation and comparison of the small sample powers of the various test procedures proposed in this paper have been given.
      0 references

      Identifiers