The performance of delta check methods

Abstract

The percentage of mislabeled specimens detected (true-positive rate) and the percentage of correctly labeled specimens misidentified (false-positive rate) were computed for three previously proposed delta check methods and two linear discriminant functions. The true-positive rate was computed from a set of pairs of specimens, each having one member replaced by a member from another pair chosen at random. The relationship between true-positive and false-positive rates was similar among the delta check methods tested, indicating equal performance for all of them over the range of false-positive rates of interest. At a practical false-positive operating rate of about 5%, delta check methods detect only about 50% of mislabeled specimens; even if the actual mislabeling rate is moderate (e.g. 1%), only about 10% of specimens flagged by a delta check will actually have been mislabeled.