Can't see the forest for the trees? Statistical considerations for disease macroecology.

Abstract

Disease macroecology relies on large, complex datasets to understand the biotic and abiotic factors shaping parasite distributions and emerging infectious disease risk. These datasets span local to global host-parasite interactions and often integrate diverse host or parasite traits across evolutionary histories. Selecting appropriate statistical approaches requires first asking key questions about the study system: How well is the system understood? How important is predictor accuracy? How much bias is present in the data? We outline three broad analytical pathways commonly used in disease macroecology (frequentist models, Bayesian approaches, and machine learning) and link them to typical data contexts, providing R coding examples using current packages. We emphasise that these pathways are intended as flexible guidelines rather than prescriptive frameworks. Although more advanced approaches have grown in popularity over the last three decades, traditional significance tests remain widely used. While these tests remain appropriate for well-controlled or small-scale experimental systems, they are often unsuitable for complex ecological datasets, where violations of assumptions, hierarchical structure, and sampling biases can lead to unstable or difficult-to-reproduce inferences. We advocate for reporting biologically meaningful predictors, effect sizes, measures of uncertainty, and transparent analytical workflows. By explicitly linking common data challenges in disease macroecology to their consequences for model choice and inference, we encourage more reproducible, transparent, and context-appropriate statistical practices for understanding and forecasting parasite distributions and emerging disease risks.