Replication in AP Statistics

In AP Statistics, replication means assigning each treatment to enough experimental units so that differences caused by the treatment can be told apart from chance variation between units. It is one of the four elements of a well-designed experiment in the CED (1.13.A.1.iii).

What is Replication?

Replication is one of the four required elements of a well-designed experiment in CED Topic 1.13 (EK 1.13.A.1.iii). The other three are comparison of at least two treatments, random assignment, and control of extraneous variation. In AP Stats, replication means using more than one experimental unit per treatment, and ideally a lot more. If you test a new fertilizer on one plant and the old fertilizer on another, any difference could just be because one plant was healthier to begin with. Put each fertilizer on 30 plants instead, and natural plant-to-plant differences start to average out.

Here's the intuitive version. Replication is how an experiment drowns out the noise. Every group of experimental units varies on its own. With only a few units per group, that natural variation can look like a treatment effect, or it can hide a real one. With enough units, a true treatment effect stands out from the chance variation. You'll also hear "replication" used for repeating an entire study with new subjects or in a new setting. That's a real scientific practice, but on the AP exam the core meaning is enough units within each treatment group.

Why Replication matters in AP® Statistics

Replication lives in Unit 1: Exploring One-Variable Data and Collecting Data, Topic 1.13 Experimental Design. It directly supports AP Stats 1.13.A (identify elements of a well-designed experiment), where it appears by name in the essential knowledge. It also feeds AP Stats 1.13.C (justify the appropriateness of an experimental design) and AP Stats 1.13.D (justify conclusions from a well-designed experiment). A cause-and-effect conclusion depends on random assignment, but random assignment only works well when there are enough units for the randomization to balance out the groups. Replication is also the bridge to inference. It's the reason later units can ask whether a difference is too big to be explained by chance alone.

How Replication connects across the course

Randomization (Unit 1)

Random assignment and replication work as a team. Randomly assigning 2 subjects per group won't reliably balance anything, because one group could easily end up with all the healthy ones. Randomly assigning 50 per group makes the groups roughly comparable on lurking variables. Randomization spreads out confounding, and replication gives it enough units to actually work.

Sample Size and Variability (Units 1, 3, 4)

Replication is basically sample size inside an experiment. The same idea shows up later in inference. Larger samples produce smaller standard errors for proportions (Unit 3) and means (Unit 4). More units per treatment means less chance variation in the group results, so real differences are easier to spot.

Statistically Significant (Units 3-4)

A result is statistically significant when it's unlikely to have happened by chance alone. Replication is what makes that judgment possible. Without enough units, you can't estimate how much chance variation to expect, so you can't tell a real treatment effect from random noise.

Matched Pairs Design (Units 1, 4)

In a matched pairs design, replication means using many pairs (or many subjects who each get both treatments), not just one. The analysis of those paired differences shows up again with paired data for means in Unit 4, and it only has power when there are enough pairs.

Is Replication on the AP® Statistics exam?

Replication shows up most often in multiple-choice questions that ask you to name or explain an element of experimental design. Typical stems ask which statement best explains why replication is essential, or which scenario violates the principle of replication. A violation might be a study that tests a new teaching method on just one class. Other questions mix replication with blocking, randomization, and control, asking which combination of design elements best addresses validity concerns. Watch for distractors. Randomizing the order in which a subject tastes chocolates is randomization, not replication.

On free-response questions, replication usually comes up when you describe or critique an experimental design. Released FRQs such as 2021 Q2 (a year-long investigation of walking and cholesterol levels in adults) and 2018 Q4 (recovery after ACL surgery) set up experimental scenarios. In those questions you're expected to apply the principles of good design. If you explain why a design is appropriate or why a conclusion is justified, mention that each treatment is applied to many experimental units so chance variation can be separated from the treatment effect. Tie it to context. For example, write "assigning 50 adults to each walking program," not just "use replication."

Replication vs Randomization

Both are principles of experimental design, and they get swapped on MCQs all the time. Randomization is about how units get assigned to treatments. Using chance prevents bias and reduces confounding. Replication is about how many units get each treatment, which needs to be enough to separate real effects from chance variation. A design can randomize perfectly and still fail at replication (one subject per group). It can also have tons of subjects but no randomization (people choose their own treatment). A strong design needs both.

Key things to remember about Replication

• In AP Statistics, replication means applying each treatment to enough experimental units, not just one or two.

• Replication is listed in the CED as one of four elements of a well-designed experiment, along with comparison, random assignment, and control.

• The purpose of replication is to make sure natural variation between units doesn't get mistaken for a treatment effect, or hide a real one.

• Randomization decides how units are assigned to treatments, while replication decides how many units each treatment gets.

• Repeating an entire study with new subjects is also called replication, but on the AP exam the main meaning is multiple units within each treatment group.

• On FRQs, describe replication in context, such as stating how many subjects receive each treatment, instead of just naming the term.

Frequently asked questions about Replication

What is replication in AP Stats?

Replication means using enough experimental units in each treatment group so that differences caused by the treatment can be distinguished from chance variation. It's one of the four elements of a well-designed experiment listed in CED Topic 1.13 (EK 1.13.A.1.iii).

Does replication mean repeating the whole experiment?

Not on the AP exam, at least not primarily. In AP Stats, replication mainly means assigning each treatment to many experimental units within one experiment. Repeating a study with new subjects is a related scientific idea, but the CED's design element is about having enough units per treatment.

What's the difference between replication and randomization?

Randomization uses chance to decide which units get which treatment, which reduces confounding. Replication makes sure each treatment goes to enough units to average out chance variation. For example, randomly assigning 1 plant per fertilizer is randomized but not replicated.

Why is replication important in an experiment?

Experimental units naturally differ from one another, so with only a few units per group, those differences can look like a treatment effect. Replication lets the natural variation average out, so you can judge whether an observed difference is statistically significant or just due to chance.

What's an example of violating replication?

A study that tests a new teaching method on one classroom and the old method on another violates replication. Any difference in test scores could come from the teacher, the time of day, or the students in that one room. Using many classrooms or students per method fixes the problem.