The detection of doubtful data: the program DOUBT.

Abstract

The problem of the detection of erroneous data has led us to define the 'doubtful patients' as patients who have a small density. Our program DOUBT computes the densities associated with each patient by using the method of n nearest neighbors. Then the program allows us to define a density limit, and chooses the patients who have a smaller density then the doubtful patients. The remaining patients may be classified by the program.