Reibnegger G, Lang T, Estelberger W
Medizinisch-Chemisches Institut und Pregl-Labor, Graz, Austria.
Clin Chim Acta. 1997 Aug 29;264(2):207-18. doi: 10.1016/s0009-8981(97)00095-8.
Information theoretical concepts can be used as most valuable tools for evaluating clinical chemical tests. They offer several advantages over the conventional test evaluation by Bayesian concepts. Although long known, information theory has failed to gain widespread acceptance among clinical chemists, probably due to the rather laborious computation of the necessary quantities. To resolve this technical problem, this paper demonstrates how commonly available computer spreadsheet programs can assist in the computation of information theoretical quantities such as the diverse entropies and the information content. Two spreadsheet versions are presented: firstly, a working sheet is developed for information theoretical evaluation of a two-by-two contingency table corresponding to a dichotomous diagnostic problem (e.g., diseased versus non-diseased) and a dichotomous diagnostic test outcome (e.g., normal versus abnormal), and secondly, a working sheet is shown for a more general situation with up to ten diagnostic alternatives and up to ten possible test outcomes. Typical examples for application of these tools are demonstrated.
信息理论概念可作为评估临床化学检验的极有价值的工具。与基于贝叶斯概念的传统检验评估相比，它们具有若干优势。尽管信息理论早已为人所知，但在临床化学家当中却未能得到广泛认可，这可能是因为计算所需数量颇为繁琐。为解决这一技术问题，本文展示了常见的计算机电子表格程序如何有助于计算诸如各种熵和信息含量等信息理论量。文中给出了两个电子表格版本：其一，开发了一个工作表，用于对与二分诊断问题（如患病与未患病）和二分诊断检验结果（如正常与异常）相对应的二乘二列联表进行信息理论评估；其二，展示了一个适用于多达十种诊断选项和多达十种可能检验结果的更一般情况的工作表。文中还演示了这些工具应用的典型示例。
