Three underused statistical methods in social epidemiology: multiple informant models, fractional regression, and restricted mean survival time

Abstract

We recommend 3 well-established yet underused statistical methods in social epidemiology: multiple informant models, the fractional regression model, and the restricted mean survival time. Multiple informant models improve how we identify critical windows of exposure over time. The fractional regression model addresses the inadequacies of ordinary least squares and logistic regression when dealing with fractional outcomes that are naturally proportions or rates, thereby accommodating data at the boundaries of the unit interval without requiring transformations. The restricted mean survival time offers a robust alternative to the hazard ratio in the presence of nonproportional hazards, providing an interpretable summary of treatment effects over time that is not dependent on the proportional hazards assumption. We illustrate the utility of each method using simulated case examples. These methodologies enrich the analytical toolbox of social epidemiologists, offering refined approaches to unraveling the complexities of social determinants of health inequities. This article is part of a Special Collection on Methods in Social Epidemiology.