POSCON — a Decision-Support System in Diagnosis and Prognosis Based on a Statistical Approach

Summary

It is natural to consider the probability that an individual, given its characteristics, belongs to a certain population. Such individual-dependent probabilities can often be defined as posterior probabilities. Basic assumptions are that k mutually exclusive populations are involved and that the individual belongs to one of these k populations or, more precisely, that it is a random drawing from one of them. The posterior probabilities can then be defined by means of Bayes’ theorem. To obtain numerical values for the thus defined individual-dependent probabilities we not only need the scores for the characteristics of the individual but also numerical values for the k prior probabilities and for the values of the k densities in the score vector x. Thee values are usually based on some estimation procedure. Accordingly there are uncertainties involved. A nice way to express uncertainties is by means of confidence intervals. The interactive program POSCON (compute POSterior CONfidence intervals) has been written for it.