Abstract
In an earlier paper by Burdick and Rayens a model was developed which uses discriminant analysis to construct estimates of linear mixing proportions. In this paper a method is proposed for identifying potentially unreliable estimates when using the Burdick/Rayens model. It is also shown to be useful in detecting irregularities in the sample data. The methodology is based on the lengths of residual vectors and is demonstrated within the context of polychlorinated biphenyl mixtures.

This publication has 1 reference indexed in Scilit: