Methodical Extensions

Abstract

There are sets of data without a clear separation of distinct patterns. This can be caused either by a high degree of complexity or insufficient homogeneity of the applied unit group . In the former situation, the discrimination threshold is adjusted in a way which ascertains that the sum of the error rates in the groups to be distinguished can be minimized. In the latter situation, the unit group is scrutinized regarding homogeneity, refined, and tried again. The latter is also used as a method to begin with if only one set of data exists to investigate if the data set is composed of hidden patterns. Thus, historic or unique data can be analyzed retrospectively to gain additional information . In the case of great amount of data, the individual Mahalanobis distances can be condensed into multiples to reduce data complexity at the risk of less accuracy. If patterns change gradually, for instance as a function of time, Mahalanobis data can be treated as a dynamic system . For therapeutic purposes, the SNR index is calculated individually for each patient to figure out the most efficient therapy with respect to time lapse.