Summary
Interrogates construct validity and measurement, arguing that every measured variable encodes a theoretical commitment and contrasting biomedical, social-determinants, fundamental-cause, ecosocial, and health-equity frameworks through a worked diabetes example. Uses the CES-D depression scale and self-rated health to show differential item functioning across cultural and socioeconomic groups, plus the dangers of treating Likert ordinal data as interval and the regression dilution that attenuates poorly measured exposures. Introduces directed acyclic graphs with confounder, mediator, and collider structures, working the Hollywood talent-attractiveness paradox and tracing the consequences of conditioning on a collider, before previewing residual confounding, reverse causation, and simultaneity bias.