Measurement Error

This chapter explores the challenges epidemiologists face when inferring disease patterns from noisy or indirect measurements of risk factors. Such measurement errors arise from various sources, such as reliance on self-reported data and intrinsic biological variation. The focus here is on the impact on regression models fitted with imperfect measurements of predictor variables. Specifically, the chapter examines statistical issues in modeling response variables in relation to mismeasured and error-free predictors. The emphasis is on measurement error for continuous predictor variables rather than misclassification of categorical predictors. A case-control study subject to covariate measurement error is used to elucidate the concepts. The chapter is organized into three sections: basic concepts and models of measurement error, planning studies in the presence of measurement error, and methods for analyzing data with measurement error. It is intended as an introduction, with references to the extensive literature offering more comprehensive descriptions of measurement error models, including linear and non-linear approaches, Bayesian perspectives, and misclassification issues. In depth review articles in the epidemiological literature also address measurement error, reflecting its significance in population-based research.