基于分层四格表资料的边际结构模型分析方法及其 SAS 实现

摘要

目的：结合病例对照研究中的分层四格表资料，介绍边际结构模型的思想，探索边际结构模型的分析方法。方法利用“处理倒数概率加权”的方法，使“原初人群”经过“补实”后，获得“虚拟人群”，经过Mantel-Haenszel 估计，计算ORMH值，并计算未经分层的cOR，比较ORMH与cOR的差异。结果边际结构模型的分析结果与Mantel-Haenszel估计的分析结果基本一致。结论边际结构模型为我们解决上述问题提供了一种全新的思路。%Objective Combined with the Four-fold tables with confounders in the case-control study, introduce the idea of marginal structural models and explore the analytic method of marginal structural models.Methods Using the method of″In-verse-Probability-of-Treatment Weighting″, so that we can obtain a″Pseudo population″based on the″original people″through the″Counterfactual″, using the Mantel-Haenszel estimate in the″Pseudo-population″, calculated the value of ORMH , compared ORMH with cOR.Results The result of marginal structural models and the result of Mantel-Haenszel estimate are basically consistent. Conclusion Marginal structural provided a new way of thinking to solve the above problem.