回帰分析を用いた疫学データの解析　基本編（3）

抄録

We have covered the use and interpretation of regression analysis for risk factor analysis in the last two articles. This article steps back and introduces a very traditional and basic analysis to explain the fundamental difference between confounding and interaction/effect modification. I explain the concept of Mantel-Haenszel (MH) estimator and its analytical method using a mock data. While the readers may not use MH estimator frequently, I hope this article explains the concept of stratification and meaning of variable selections in regression analyses.