C37. Conditional versus marginal effects

Topic: Marginal models, and two responses at once · more questions on this topic
‹ C36 C38 ›

Why is a logistic GLMM coefficient not generally the same as a population-average logistic coefficient?

Worked answer (click to show)

Definitions/notation. A logistic GLMM coefficient describes a log-odds change conditional on a fixed random-effect value; a population-average coefficient describes the marginal mean after averaging over random effects. Because the inverse-logit is nonlinear, averaging conditional probabilities generally does not preserve the same logistic coefficients.