18: Confounding and Causal Structure

Author

Affiliation

Smit, A. J.

University of the Western Cape

1 Introduction

This chapter extends the principles of model specification into the domain of causal inference. We move from simply selecting variables to understanding their causal roles. A clear grasp of confounding is essential for avoiding spurious conclusions and for making stronger, more defensible claims about the processes driving ecological patterns.

2 Key Concepts

• Confounders vs. Mediators: Learning the critical difference between a variable that causes both a predictor and an outcome (a confounder) and a variable that lies on the causal pathway between them (a mediator).
• Spurious Relationships: Identifying relationships that are statistically significant but not causal, often due to a hidden common cause.
• Directed Acyclic Graphs (DAGs): Using simple graphical models (DAGs) to map out proposed causal relationships and to identify potential confounding variables before fitting a statistical model.
• Link to Collinearity: Clarifying how confounding is a causal concept, distinct from collinearity, which is a statistical property of the data.

This chapter is a placeholder. Content will be developed further.