21: Measurement Error and Proxies

Author

Affiliation

Smit, A. J.

University of the Western Cape

1 Introduction

Our models assume that predictor variables are measured without error, but in ecology, this is rarely true. This chapter explores the consequences of measurement error and the related concept of using proxy variables, consolidating ideas first touched upon in the chapters on collinearity and confounding.

2 Key Concepts

• Measurement Error in Predictors: Understanding how random error in the measurement of a predictor variable affects model estimates.
• Attenuation Bias: Recognizing that measurement error in a predictor typically biases its estimated coefficient toward zero, making it harder to detect an effect.
• Proxy vs. Mechanistic Variables: Discussing the use of easily measured proxy variables (e.g., altitude) to stand in for the true, mechanistic variable of interest (e.g., temperature, radiation stress), and the inferential challenges this poses.

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