日常诊断——对贝叶斯模型的批判

Everyday diagnostics--a critique of the Bayesian model.

作者信息

Jonson N E

机构信息

Surgical Department, Central Hospital, Kristianstad, Sweden.

出版信息

Med Hypotheses. 1991 Apr;34(4):289-95. doi: 10.1016/0306-9877(91)90044-y.

DOI:10.1016/0306-9877(91)90044-y

PMID:1865835

Abstract

In recent years Bayesian probability calculus and Bayesian decision procedures have been recommended for use in clinical medicine. The author investigate everyday diagnostics asking if it takes place in a reality that meets the conditions of the method. He finds it does not. Above all we lack that strict randomness essential for probability calculus and, further, evaluation of utility easily becomes disputable, not to say unethical. The measures constructed for neutralizing these discrepancies between model and reality demand great resources. Yet, they are not enough to permit the clinician to refrain from supervising the consequences of his decisions as carefully as he has always done. A different method, a different model is wanted.

摘要

近年来，贝叶斯概率演算和贝叶斯决策程序已被推荐用于临床医学。作者研究了日常诊断，询问其是否发生在符合该方法条件的现实中。他发现并非如此。首先，我们缺乏概率演算所必需的严格随机性，此外，效用评估很容易变得有争议，甚至可以说是不道德的。为消除模型与现实之间的这些差异而构建的措施需要大量资源。然而，这些资源还不足以让临床医生像以往那样不再仔细监督其决策的后果。我们需要一种不同的方法、一个不同的模型。