On a classification rule for multiple measurements (Q3747529)

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 3984358
Language Label Description Also known as
default for all languages
No label defined
    English
    On a classification rule for multiple measurements
    scientific article; zbMATH DE number 3984358

      Statements

      On a classification rule for multiple measurements (English)
      0 references
      0 references
      1986
      0 references
      multiple measurements
      0 references
      decision-theoretic model of the classification problem
      0 references
      optimum classification procedure
      0 references
      univariate classification procedure
      0 references
      maximum-likelihood criterion
      0 references
      conditional probability of misclassification
      0 references
      chi-squared distribution
      0 references
      non-integer number of degrees-of-freedom
      0 references
      A univariate classification procedure, based on the maximum-likelihood criterion, is derived when each individual observation has been made in replication. The classification rule R is: Classify X to \(\pi_ i\) \((i=1,2)\) iff \(aS^ 2+b\bar X^ 2-2c\bar X-d\geq 0\), where the constants a, b, c and d depend on the parameters of the model: \[ Y_{ijk}=\mu_ i+I_{ij}+E_{ijk},\quad j=1,...,n_ i,\quad k=1,...,n, \] \(I_{ij}\) and \(E_{ijk}\) are distributed independently as \(N(0,\nu^ 2_ i)\) and \(N(0,\tau^ 2_ i).\) The distribution of the classification statistic is found to be the weighted difference of two independent noncentral chi-squared variates plus a constant. Formulae for the conditional probability of misclassification are given, and an approximation as a chi-squared distribution with a non-integer number of degrees-of-freedom is proposed. Two special cases of the problem are studied: (i) \(\tau^ 2_ 1=\tau^ 2_ 2\) and (ii) \(\tau^ 2_ 1=\tau^ 2_ 2\) and \(\nu^ 2_ 1=\nu^ 2_ 2\). Further, the problem is considered when the parameters of the underlying distributions are unknown and is extended to more than two populations, respectively.
      0 references

      Identifiers