Small effects: Subgroup analysis and interaction

Abstract

It is common in epidemiologic studies to investigate relationships between exposure and disease, both in the entire study population and in strata or subgroups. Also in situations where the effect in the entire study population, sometimes referred to as the main effect, is absent or small, is it common practice to continue with subgroup analysis. One rationale for looking at effects in subgroups is that there may be segments of the population in which people have an increased susceptibility and in which an effect of the exposure may be strong or at least easier to detect. There may also be methodological differences across segments of the population that make it easier to detect relations in some parts of the population than in others.