Vol: 30 Page: 246

Authors:
Wiesław Pasewicz

Title:
 On probability of misclassification for zero-mean discrimination with special covariance matrix structure

Language: English

Keywords:
quadratic discriminant function
probability of misclassification
special covariance matrix

Summary:
Let us assume that in population πi random vector X has p-dimensional normal distribution with zero-mean vector and covariance matrix of the form Σi=σi2[(1-ρ)I+ρE]; i=1,2, where σi2 is the variance, ρ is the common correlation coefficient, I is the (pxp) identity matrix and E is the (pxp) matrix with all elements equal to 1.
In the paper we find the misclassification probabilities for the quadratic discriminant function when the matrix Σi, is known. The case when the parameter ρ is known but σi2 is unknown is considered as well.