Course
Foundational60 min
Measurement Error Basics
Slide 1 of 7
What is Measurement Error?
The Third Major Bias
After confounding and selection bias, measurement error (also called information bias) is the third major threat to valid causal inference. It occurs when our measured values differ from the true values we intend to capture.
Measurement Error Defined
Measurement error occurs when the recorded value of a variable differs from its true value. For continuous variables, this is typically modeled as: $X^* = X + \epsilon$, where $X^*$ is the measured value, $X$ is the truth, and $\epsilon$ is the error.
Where Measurement Error Can Occur
| Variable | Consequence of Error | Example |
|---|---|---|
| Exposure | Biased treatment effect | Self-reported diet vs. biomarkers |
| Outcome | Biased effect estimate | Undiagnosed disease |
| Confounders | Residual confounding | Crude vs. detailed smoking measure |
Terminology
For continuous variables: We typically use "measurement error"
For categorical variables: We use "misclassification"
- Binary variable misclassified as its opposite
- Multi-category variable placed in wrong category
Why This Matters
Measurement error is ubiquitous in epidemiology:
- Self-reported exposures are often inaccurate
- Disease diagnoses can be wrong
- Biomarkers have laboratory variation
- Administrative data have coding errors
From Quantitative Bias Analysis
"Measurement error is present in virtually every epidemiologic study. The question is not whether it exists, but how large it is and what impact it has on the study results."
Source: Quantitative Bias Analysis (Lash et al.), Chapter 6
Use ← → arrow keys to navigate