ABSTRACT
Chapter 4 presents effect-measure modification and causal interaction. It illustrates effect-measure modification with a subgroup analysis in a randomized clinical trial, and it relates effect-measure modification to statistical interaction, particularly within linear, loglinear, and logistic models. It presents results on a sufficient condition for qualitative agreement of four effect measures in modification, together with a Monte-Carlo study and real data example bearing on pairwise qualitative agreement. It distinguishes between causal interaction and statistical interaction, and presents causal interaction in terms of the potential outcomes framework for two causes that may synergize. It enumerates sixteen causal types, defines monotonicity of those types, and presents a sufficient condition for two causes to synergize. Examples and R code are also provided.
