Between-Subject Heterogeneity: Random-Effects Models for Education and Mortality

This chapter motivates random-effects modelling of education-related mortality, contrasting it with fixed effects and clarifying the random-effects assumption of no correlation between random effects and covariates. We fit random-intercept logistic models in SAS PROC GLIMMIX with Laplace approximation, quantify the intraclass correlation, and consider extensions with a random age slope and Education × Year interaction. Exploratory cross-tabulations show modest yet significant bivariate gradients by education. We convert predicted probabilities into life-table metrics and policy-facing cost-effectiveness. A health-literacy programme is highly cost-effective, outperforming vouchers and bursaries. Substantively, age dominates, and education attenuates once covariates and heterogeneity are accounted for, guiding policy translation.