Matched Pairs

TL;DR

Compares two related measurements taken on the same subjects (also called matched samples or paired samples).
Controls for individual differences by using within-subject comparisons (e.g., before vs. after).
Commonly analyzed with the paired t-test or the Wilcoxon signed-rank test.

Definition

Matched pairs, also known as matched samples or paired samples, are a type of statistical analysis where two related measurements are taken from the same subjects and then compared to determine if there is a significant difference between the two.

Explanation

In matched-pairs designs each subject provides a pair of measurements that are directly related (for example, a measurement before and after an intervention). Comparing these paired measurements controls for individual differences across subjects, allowing a more accurate assessment of whether a real change or difference exists. Statistical tests for matched pairs compare the differences within each pair to assess significance.

Examples

Blood pressure medication (before and after)

A researcher takes blood pressure measurements from a group of patients before and after they have been given a medication. The matched pairs are the blood pressure readings taken from each individual patient before and after taking the medication.

Teaching method (before and after)

A researcher takes test scores from a group of students before and after implementing a new teaching method. The matched pairs are the test scores of each individual student before and after the new teaching method was implemented.

Notes or pitfalls

Matched pairs control for individual differences (for example, some patients may have naturally lower blood pressure than others), making within-subject comparisons more accurate than independent-sample comparisons in these settings.

Matched samples (synonym)
Paired samples (synonym)
Paired t-test
Wilcoxon signed-rank test