Statistical analysis-measurement error

Timo B. Brakenhoff*, Maarten van Smeden, Daniel L. Oberski

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

Abstract

An important aspect of data quality when conducting clinical analyses using real-world data is how variables in the data have been recorded or measured. The discrepancy between an observed value and the true value is called measurement error (also known as noise in the artificial intelligence and machine learning literature) and can have consequences for your analyses in all kinds of contexts. To properly assess the potential impact of measurement error it is essential to understand the relationship between the true and observed variables as well as the goal of the analysis and how it will be implemented in practice. Commonly, measurement error is distinguished as being classical, Berkson, systematic and/or differential. While it is clear that measurement error can have far-reaching consequences on analyses, the effect can differ depending on whether analyses are descriptive, explanatory or predictive. Validation studies can inform the estimation and characterization of measurement error as well as provide crucial information for correction methods that are available in several statistical programming languages such as SAS, R and Python.

Original languageEnglish
Title of host publicationClinical Applications of Artificial Intelligence in Real-World Data
EditorsFolkert W. Asselbergs, Spiros Denaxas, Daniel L. Oberski, Jason H. Moore
PublisherSpringer
Pages97-108
Number of pages12
Edition1
ISBN (Electronic)9783031366789
ISBN (Print)9783031366772
DOIs
Publication statusPublished - 5 Nov 2023

Keywords

  • Bias
  • Correction
  • Estimation
  • Measurement error
  • Misclassification
  • Modelling
  • Noise

Cite this