A discrimination value of serum lipid level on the basis of healthy and diseased persons

Abstract

We have established a discrimination value that classifies serum lipid levels, that is, total cholesterol, phospholipid, triglyceride, and beta-lipoprotein, into normal or abnormal state. The analysis was made using the linear discriminant function based on 720 normal subjects and 294 patients with coronary heart disease and/or diabetes mellitus. Basically, the function denoted as f is a linear combination of the predictor variables x(1),x(2),...,x(n) weighted by w(0),w(1),w(2),...,w(n). First, we obtained a function from the training set that consists of age-matched 204 normals and 204 patients. The predictor variables used are four serum constituents and age. Next, we tested the cross-validation of this function by the remainder of the samples (test set). Finally, we obtained a discrimination value, by which 96.1% of the training set and 96.3% of the test set was correctly classified.