关于混杂效应、交互效应的探讨

The Exploration of Confounding and Interaction Effect

目的:探讨在实际处理数据时对混杂效应、交互效应的处理方法。方法:通过具体实例说明混杂效应、交互效应的概念及正确应用。结果:判断变量间是否存在交互效应,需要在模型中纳入变量间的乘积项并通过统计学检验来评价;判断某变量是否为混杂因素,需比较模型中没有纳入该变量时得到的粗估计值与该变量纳入模型时得到的校正估计值的差别是否具有实际意义的不同,而不是通过统计学检验来评价;结论:在对数据进行统计分析时,当某变量可能与其它变量存在交互作用,同时又考虑其可能为混杂因素时,应先考虑其是否存在可能的交互作用,因为研究因素的效应随其他变量的取值不同而变化,如交互作用无统计学意义,进一步评价其是否为混杂因素。

Objective： To explore the application of confounding and interaction effect in data analysis.Method： The definition and correct application of confounding and interaction are illustrated by specific examples.Results： Interaction effect exists when the relationship of interest differs at different levels of extraneous variables and is evaluated by using statistical test;confounding is present when the effect of interest differs depending on whether an extraneous variable is ignored or included in the data analysis,any clinically important change in the estimated coefficient for the risk factor between the crude and adjusted estimates suggests that the covariate is a confounder and should be included in the model,regardless of the statistical significance of its estimated coefficient.Conclusion： If strong interaction is found,interaction should be assessed before confounding is assessed,interaction takes precedence over confounding,since the estimate of the effect of the risk factor depends on the specific value of the covariate,confounding is recommended only when there is no meaningful interaction.