Bias in statistics refers to a systematic error that leads to inaccurate or misleading results. Statistical bias describes any aspect that creates a difference between an expected value and the true value of a population parameter being estimated. It can also be thought of as the consistent underestimation or overestimation of a population parameter value. Bias can arise in study design, data collection, or analysis, causing estimates to consistently deviate from the true value. Reducing bias is essential to producing valid and reliable conclusions.
Common types of statistical bias include the following:
-
Selection bias: occurs when the sample is not representative of the population.
-
Measurement bias: arises from errors or inaccuracies in how data are measured or recorded.
- Sampling bias: occurs when the method of selecting participants or data points leads to a sample that is not representative of the target population, resulting in systematic errors in estimates.
-
Model misspecification bias: occurs when the statistical model used does not correctly represent the underlying relationship between variables, such as leaving out important predictors or using the wrong functional form.
Each of these types are broad categories containing multiple subtypes of statistical bias.
For example, suppose a researcher studies the average income in a city but only surveys people leaving luxury stores. This introduces selection bias, as the sample overrepresents high-income individuals. As a result, the data will overestimate the true average income of the city, leading the researcher to incorrectly conclude that the city is wealthier overall than it actually is.