Introduction to Bias Analysis

Slide 1 of 7

Why Quantitative Bias Analysis?

Beyond Random Error

Standard statistical analysis focuses on random error—the variability due to sampling from a population. Confidence intervals and p-values address this uncertainty. But they completely ignore systematic error (bias), which is often the larger threat to validity.

Quantitative Bias Analysis (QBA)

A collection of methods to quantify the potential impact of systematic errors (confounding, selection bias, measurement error) on study results. QBA does not claim to eliminate bias but helps understand how robust conclusions are to plausible biases.

The Two Types of Error

Error Type	Source	Addressed By	Reducible By
Random	Sampling variability	CI, p-values	Larger sample size
Systematic	Design, measurement	Bias analysis	Better design, QBA

The Problem with Standard Analysis

Consider a study reporting: RR = 2.0, 95% CI (1.5, 2.7), p < 0.001

Standard interpretation: "Strong evidence of an effect"

But what about:

Unmeasured confounding?
Selection into the study?
Exposure misclassification?

These sources of bias are invisible to confidence intervals.

The Precision Paradox

Larger sample sizes make confidence intervals narrower, increasing apparent precision. But systematic errors remain unchanged—they may even dominate as random error shrinks. A very precise but biased estimate is not useful.

From Quantitative Bias Analysis

"Every epidemiologic study is susceptible to systematic errors. Ignoring them does not make them go away—it only makes us ignorant of their potential impact."

Source: Quantitative Bias Analysis (Lash et al.), Chapter 1

Use ← → arrow keys to navigate