← Back to course

Matched Pairs & Why Pairing Beats Two Independent Groups

You’ll be able to

What pairing does

A matched-pairs design compares two conditions within the same subject, or within pairs of subjects deliberately matched on characteristics likely to affect the response. Each subject receives both treatments, or each pair splits them at random. The analysis then works on the differences within each pair — one number per pair rather than two independent groups of numbers. That single structural choice is what makes the design powerful.

Why it is more sensitive

With two independent groups, the variation you are trying to detect competes against all the variation between subjects — people differ enormously in reaction time, blood pressure, test-taking ability. Pairing removes that source entirely, because each subject serves as their own control and any characteristic they carry into both conditions cancels in the difference. What remains is the treatment effect plus much smaller within-subject noise, so a real effect is far easier to detect at the same sample size. This is the statistical meaning of a "more powerful" design.

The analysis follows the design

A matched-pairs experiment is analyzed with a one-sample t procedure on the differences, not a two-sample t procedure. Compute d = (value under condition A) − (value under condition B) for each pair, then test whether the mean difference is zero or build a confidence interval for it. Applying a two-sample procedure to paired data throws away the pairing, inflates the standard error, and is a standard way to lose points even when the arithmetic afterward is flawless.

On the exam

The diagnostic question is simple: is there a natural reason to link one observation in the first group with one specific observation in the second? Before-and-after on the same person, left hand and right hand, twins, plots of land split in two — all paired. Two separate randomly assigned groups of different people — not paired.

Worked example

Twelve runners each complete a course wearing shoe A and, on a separate day in random order, shoe B. Identify the design, state the correct procedure, and explain what pairing accomplishes here.

  1. 1.Each runner experiences both conditions, so the runners are the pairs — this is a matched-pairs design.
  2. 2.Compute one difference per runner, time in shoe A minus time in shoe B, giving twelve differences.
  3. 3.Analyze with a one-sample t test or interval on those twelve differences, with 11 degrees of freedom.
  4. 4.Randomizing the order across days controls for learning the course and for day-to-day conditions, which would otherwise be confounded with the shoe.
Answer: This is a matched-pairs design analyzed with a one-sample t procedure on the twelve differences, with df = 11. Pairing removes the very large runner-to-runner variation in speed — the fastest and slowest runners in the study may differ by minutes while the shoe effect may be seconds — so the shoe difference is measured against within-runner noise instead of being buried in between-runner spread.
Checkpoint

Blood pressure is measured on 30 patients before and after a drug regimen. The appropriate inference procedure is —

Checkpoint

The chief statistical advantage of a matched-pairs design over two independent groups is that it —

Answer the 2 checkpoints as you read.

Sign in to save your progress