The utilities and pitfalls of stratified analysis in challenging situations

Abstract

Stratified analysis is commonly used to control confounding factors in evaluating treatment effect at both design and analysis stages. Motivated by a real-life experience, this article illustrates the utilities and pitfalls of stratified analysis through continuous and binary endpoints by comparing some typical analysis methodologies. Stratified analyses in some challenging situations, including propensity score stratification are discussed.

Keywords: Farrington-Manning; Simpson's paradox; propensity score; stratified analysis.