医療データのreductionに関する統計論的アプローチ

抄録

Questionnaires used generally at present are so numerous that they are troublesome both for patients and physicians. To solve this problem, the following method is introduced.
The feature evaluation function is determined on the basis of the idea of minimizing the upper bound of mean error probability in Neyman-Pearson Decision Rule. According to this determined evaluation function, each inquiry is evaluated and ranked in its degree of contribution to discriminating diseases. That is to say, m-number of inquiries are to be selected from among n-number of inquiries, thus reducing them from n to m.
The problem is how to deal with false positive and false negative in case of discriminating diseases. A decision rule capable of controlling the ratio between the two was proposed.
In order to evaluate above-mentioned two theories, the rule was applied to discriminating healthy persons from patients with heart-diseases. The result was, as far as these theories were concerned, satisfactory.