The Random-Effects Model in Discriminant Analysis

From MaRDI portal





This paper describes various characteristics of the classical linear discriminant model in which each mean vector \(\mu_ i\) (from population i) has been randomly selected from some population in advance of the experiment. Once the mean vectors have been selected the problem of linear discriminant analysis is to classify vectors into one of the (now) fixed populations. In different experiments there are different sets of \(\mu_ i\) all drawn independently from the same parent population. It is assumed that the elements within any randomly selected population are normally distributed with mean vector \(\mu\) and covariance matrix \(\Sigma\), and that over different populations \(\mu\) has a normal distribution with known mean vector and covariance matrix. The distribution of the population and sample-based Mahalanobis distances between two different populations and those between a randomly selected population and an observation are given. Bounds are also derived for the expected probabilities of misclassification (for both, two and more population problems) for the case where either the population-based or the sample-based linear discriminant functions are used.











This page was built for publication: The Random-Effects Model in Discriminant Analysis

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3324857)