Why your DAG should probably incorporate sampling and measurement processes

Area of Science:

• Ecology
• Entomology
• Causal Inference

Background:

• Directed Acyclic Graphs (DAGs) are increasingly used by insect scientists to visualize pre-data causal assumptions.
• Current DAG applications often overlook the crucial role of observation processes (sampling and measurement) in interpreting associations.
• This oversight can limit the ability to draw robust causal conclusions from observed data.

Purpose of the Study:

• To explain how insect scientists can integrate assumptions about sampling and measurement into DAGs.
• To demonstrate that DAGs can serve as a general framework for displaying assumptions for various inferential goals.
• To highlight the necessity of explicit sampling and measurement assumptions for holistic causal reasoning.

Main Methods:

• Drawing on interdisciplinary literature to inform the application of DAGs in entomology.
• Developing a conceptual framework for incorporating observation process assumptions within DAGs.
• Illustrating how explicit assumptions facilitate causal interpretation of observed associations.

Main Results:

• Observed associations in insect science are contingent upon both causal structure and observation process assumptions.
• Explicitly including sampling and measurement assumptions in DAGs enhances the rigor of causal inference.
• DAGs incorporating observation processes are valuable for descriptive and predictive questions, not solely causal inference.

Conclusions:

• DAGs offer a powerful, generalizable framework for visualizing and reasoning about scientific assumptions in entomology.
• Integrating sampling and measurement assumptions into DAGs is essential for robust causal interpretation and broader inferential goals.
• This expanded use of DAGs promotes more comprehensive and transparent scientific reasoning in insect science.