Cultural landscape clustering and zoning of traditional rural settlements.

In 2016, environmental hazards, including air pollution, extreme weather, and harmful chemical exposures, accounted for an estimated 24% of global deaths, or approximately 14 million annually. In the United States, ten climate-sensitive events in 2012 resulted in 10.0 billion dollars in health-related costs (2018 dollars). Although these threats are widespread, vulnerable groups, particularly those with limited capacity to anticipate, cope with, or recover from harm, face disproportionate risks. Understanding population heterogeneity in environmental health risks is therefore essential for designing targeted and equitable policies and interventions. With survey-based population studies becoming increasingly rich and multifaceted, researchers are now better equipped with data to capture a wide range of factors that may differentially influence individual health. Despite these advances, characterizing population heterogeneity remains challenging, as the relevant factors are often complex, and not easily reducible to a single or limited set of measurable variables. While traditional regression-based approaches offer a straightforward and interpretable way to quantify heterogeneity, they are often inadequate in capturing the multidimensional nature of these relationships. Moderated multiple regression (MMR) extends linear regression, allowing researchers to use familiar tools, such as hypothesis testing, to evaluate the statistical significance and magnitude of moderation effects. However, MMR has several limitations: it is prone to structural multicollinearity due to correlations between interaction terms and their components; its statistical power decreases as additional interaction terms are added; and it requires strong prior hypotheses to determine the number and order of interaction terms to include. Consequently, practitioners struggle to model and uncover complex heterogeneity when moderation effect is distributed across multiple variables.