In AP Statistics, matching is an experimental design technique that pairs experimental units with similar characteristics (or uses each subject twice), then randomly assigns treatments within each pair, reducing variability from confounding variables so treatment effects are easier to detect.
Matching is a way of controlling confounding variables before randomization ever happens. Instead of tossing all your experimental units into one big random lottery, you first group them into pairs that are as similar as possible on variables you think might affect the response (age, fitness level, soil quality, whatever matters). Then, within each pair, you randomly assign one unit to each treatment. Because the two units in a pair are so alike, any difference in their responses is more likely caused by the treatment itself.
The most common version on the AP exam is the matched pairs design, and it comes in two flavors. Either you pair up two similar subjects and split the treatments between them, or each subject serves as their own pair by receiving both treatments (with the order randomized). Think of it this way. A completely randomized design hopes random assignment balances out the noise. Matching doesn't hope, it builds the balance in directly, then randomizes what's left.
Matching lives in Topic 3.5 (Introduction to Experimental Design) in Unit 3: Collecting Data. It supports learning objective 3.5.C (compare experimental designs and methods), and it ties straight into 3.5.B, since a well-designed experiment requires control of potential confounding variables where appropriate. Matching is one of the main tools for that control. When the AP exam asks you to design an experiment or critique one, recognizing when matching beats a completely randomized design (and when it doesn't) is exactly the comparison skill 3.5.C is testing.
Keep studying AP® Statistics Unit 1
Random Assignment (Unit 3)
Matching never replaces random assignment, it works alongside it. After you form pairs, you still randomly assign treatments within each pair. Matching handles the confounders you can see; randomization handles the ones you can't.
Confounding Variable (Unit 3)
Matching exists because of confounding variables. By pairing units that are similar on a suspected confounder, you stop that variable from muddying the comparison between treatments.
Control Group (Unit 3)
In a matched pairs design, one member of each pair often gets the treatment while the other acts as the control. The pair structure makes the treatment-versus-control comparison sharper because the two units started out nearly identical.
Paired Data Inference (Unit 7)
How you design the experiment decides which inference procedure you use later. Matched pairs data gets a paired t-procedure on the differences, while independent groups (like the 60 separately assigned irrigation plots) get a two-sample procedure. Mixing these up is a classic AP error.
Matching shows up two ways. First, in MCQs that ask you to identify a design ("each of 30 participants types with both Layout A and Layout B in random order" is matched pairs, not completely randomized) or to pick the right inference procedure for paired versus independent data. Second, in design FRQs where you describe or justify an experiment. If a question gives you subjects with an obvious lurking variable, explaining a matched pairs design and how you'd randomize within pairs is often the strongest answer. Be ready for the flip side too. A common question asks when matched pairs is disadvantageous compared to a completely randomized design, such as when there's no good variable to match on, when matching is impractical, or when order effects and carryover are a problem in a repeated-measures setup. Always say the magic phrase "randomly assign within each pair," because matching without randomization earns no credit.
Random assignment is the lottery that decides who gets which treatment; matching is the sorting you do before the lottery. In a completely randomized design, randomization alone is expected to balance confounding variables across groups. In a matched design, you force the balance on known variables by pairing similar units first, then randomize within pairs. The key point is that matching includes random assignment, it doesn't replace it. A design that matches subjects but lets the researcher choose who gets which treatment is not a valid experiment.
Matching pairs experimental units with similar characteristics, then randomly assigns treatments within each pair to control known confounding variables.
A matched pairs design can use two similar subjects per pair or one subject who receives both treatments in a randomized order.
Matching does not replace random assignment; you must still randomize treatments within each pair, or the design isn't a valid experiment.
Matched pairs is disadvantageous when there's no meaningful variable to match on, when pairing is impractical, or when receiving one treatment changes a subject's response to the next one.
The design carries forward to inference: matched pairs data calls for paired procedures on the differences, while independently assigned groups call for two-sample procedures.
Matching is an experimental design technique where you pair experimental units with similar characteristics (or have each subject receive both treatments), then randomly assign treatments within each pair. It controls confounding variables so differences in the response are easier to attribute to the treatment.
No. Matched pairs only helps when you can match on a variable that actually affects the response. If there's no good matching variable, or if getting one treatment changes how a subject responds to the second (carryover effects), a completely randomized design is the better choice.
Random assignment is the random lottery for treatments; matching is the pairing you do before that lottery. Matching directly controls confounders you can identify, while randomization within each pair balances everything else. A matched design still requires random assignment within pairs.
No. One common version uses each subject as their own pair, like 30 participants each typing on Layout A and Layout B with the order randomized. The 'pair' is the two measurements on the same person, and you analyze the differences.
Ask whether each observation in one group is naturally linked to a specific observation in the other. Same subject measured twice, or deliberately paired units, means matched pairs (paired procedures on differences). Separate, independently assigned groups, like 30 plots on Method A and 30 different plots on Method B, means two independent samples.
Connect this key term to the AP exam workflow: review the course, practice questions, and check related study tools.
Review units, study guides, and course resources.
Check this vocabulary in multiple-choice context.
Apply key concepts in written AP responses.
Estimate the exam score you are working toward.
Review the highest-yield facts before practice.
Put the full course together before test day.