A biologist's guide to model selection and causal inference.

Area of Science:

• Biology
• Ecology
• Evolution
• Behavioral Science
• Epidemiology

Background:

• Biological research often involves analyzing large, complex datasets, particularly in ecology, evolution, and behavior (EEB) and epidemiology.
• EEB researchers frequently use long-term observational data to model biological processes and test causal hypotheses.
• Epidemiologists analyze similar data to understand human health determinants, often employing explicit causal directed acyclic graphs (DAGs).

Purpose of the Study:

• To review recent causal inference literature and propose an analytical workflow applicable to EEB research.
• To define four distinct analytical tasks: description, prediction, association, and causal inference.
• To focus on causal inference using DAGs to guide modeling strategies for observational data.

Main Methods:

• Review of recent causal inference literature.
• Definition of four analytical tasks: description, prediction, association, and causal inference.
• Description of an analytical workflow centered on causal directed acyclic graphs (DAGs) for observational data.

Main Results:

• Identified a key difference in analytical workflows between EEB and epidemiology regarding causal DAGs.
• Presented a framework for causal inference from observational data relevant to EEB.
• Clarified the role of DAGs in informing modeling strategies for causal inference.

Conclusions:

• The proposed analytical workflow, utilizing causal DAGs, can enhance causal inference in EEB research.
• Facilitating interdisciplinary exchange can improve the application of causal inference methods.
• This framework provides a structured approach for addressing causal questions using observational biological data.