Showing papers by "William G. Hunter published in 1979"

Discussion of: Statistical Comparison of Multiple Analytic Procedures: Application to Clinical Chemistry

Abstract: The paper by Lawton, Sylvestre, and Young addresses the important problem of the statistical comparison of several analytic methods (all for measuring the same thing) when no referee method is available. The authors propose a multivariate approach to the problem, which provides estimates of some of the components of variance for each method. The procedure is intended for situations in which fairly large data sets are available from more than two methods. We tried our hand at implementing the recommended procedure on the data set used in the paper by Carey, Wold, and Westgard [2], which was used for the illustrative example in the paper under discussion. We will call this the Carey data set. We also tried the procedure on a smaller set of constructed data. The calculations are easy to program, provided one already has programs to invert matrices and to compute eigenvalues and eigenvectors.