Bouyer J
INSERM U170: Unité de Recherches Epidémiologiques et Statistiques sur l'Environnement et la Santé, Villejuif.
Rev Epidemiol Sante Publique. 1991;39(2):183-96.
The logistic regression is used in epidemiology to study the relationships between a disease in two modalities (diseased or disease free) and risk factors Xi which may be qualitative as quantitative variables. According to this model, the probability of disease knowing Xi's values is written: [formula: see text]. The coefficients satisfy to the equation: ORi = exp(beta i) where ORi is the odds-ratio linked to the variable Xi adjusted on the other variables of the model. The second part of this paper is devoted to the tests of hypothesis in the logistic model and to the choice of the variables to be included in the model. The way of using tests to decide how to code variables is explained. Analysis strategy (choice and selection of variables) is discussed. The rational of goodness of fit tests is shown. At least, we indicate how to use the logistic model in case-controls studies and in studies with matched samples.
逻辑回归在流行病学中用于研究疾病的两种状态（患病或未患病）与可能为定性或定量变量的风险因素Xi之间的关系。根据该模型，已知Xi值时患病的概率可表示为：[公式：见原文]。系数满足方程：ORi = exp(βi)，其中ORi是与变量Xi相关的比值比，已根据模型中的其他变量进行调整。本文的第二部分致力于逻辑模型中的假设检验以及模型中要纳入变量的选择。文中解释了如何使用检验来决定变量的编码方式。讨论了分析策略（变量的选择和筛选）。展示了拟合优度检验的原理。最后，我们指出如何在病例对照研究和匹配样本研究中使用逻辑模型。
