Blocking is an experimental design technique where you group similar experimental units together (based on a variable you know affects the response), then randomly assign treatments within each block, so differences between groups don't get mistaken for treatment effects.
Blocking means sorting your experimental units into groups (blocks) based on a variable you already know affects the response, then randomly assigning treatments within each block. Think of a fertilizer experiment on a field with a moisture gradient. Wet plots will grow better than dry plots no matter what fertilizer you use. If you ignore that, the moisture differences inflate your error variance and bury the fertilizer effect. If you block by moisture level (wet plots in one block, dry plots in another) and randomize fertilizer within each block, every comparison happens between similar plots. The treatment effect becomes much easier to see.
The big idea is that blocking removes variability you can name from the variability you can't. Randomization handles unknown sources of variation by spreading them out evenly. Blocking handles a known source of variation by controlling for it directly. The two work together, never as substitutes.
Blocking is introduced with experimental design, but it pays off in Unit 7 when you start doing inference for means. Topic 7.1 and learning objective AP Stats 7.1.A focus on a core question: random variation can cause errors in statistical inference, so should you worry about error? Blocking is one of your best tools for managing that worry. By cutting down the noise from a known nuisance variable, blocking shrinks the error variance in your data. Less noise means your test is more likely to detect a real treatment effect, which is exactly what "increasing the power of the test" means. So blocking isn't just a design vocabulary word. It's the design-stage decision that makes your Unit 7 inference more trustworthy.
Keep studying AP Statistics Unit 7
Confounding Variables (Unit 3)
Blocking is the standard answer to "how do you deal with a confounding variable you know about?" If moisture, age, or car model could explain the results instead of the treatment, you block on it so it can't muddy the comparison.
Randomization (Unit 3)
Blocking and randomization are teammates with different jobs. You block on the variables you know matter, then randomize within each block to balance out everything you don't know about. A block design without randomization inside the blocks isn't a valid experiment.
Power of the Test (Units 6-7)
Blocking reduces error variance, and lower error variance means a more powerful test. This is why an MCQ about a field with a moisture gradient points to blocking. Same sample size, same treatment, but a much better chance of detecting a real effect.
Type I Error and Topic 7.1 (Unit 7)
Topic 7.1 asks whether random variation could be fooling your inference. Good blocking reduces the random variation in the first place, so the design stage and the inference stage are really two ends of the same problem.
Multiple-choice questions usually hand you a scenario with a known nuisance variable and ask which data-gathering method is appropriate. A field with a moisture gradient that "inflates the error variance" is a classic setup, and blocking is the answer because it increases the power of the test. On FRQs, blocking shows up in design questions like 2024 FRQ Q3, where researchers comparing gas mileage across four car models need a design that accounts for differences between models. To earn credit, you have to do three things: name the blocking variable, explain why you're blocking on it (it's associated with the response), and state that treatments are randomly assigned within each block. "I would use blocking" with no justification doesn't score. The justification is the whole point.
These look identical on paper (split units into similar groups, then randomize) but they live in different worlds. Stratifying is a sampling technique: you divide a population into strata and take a random sample from each to get a representative sample. Blocking is an experimental technique: you divide experimental units into blocks and randomly assign treatments within each to isolate a treatment effect. Quick check: if you're selecting people, you stratify. If you're assigning treatments, you block.
Blocking groups experimental units by a known variable that affects the response, then randomly assigns treatments within each block.
Block on what you know, randomize on what you don't. Blocking controls a named source of variation while randomization balances out unknown ones.
Blocking reduces error variance, which increases the power of the test, meaning you're more likely to detect a real treatment effect.
On FRQs, you must identify the blocking variable, explain that it's related to the response, and state that randomization happens within each block.
Blocking is for experiments and stratified sampling is for surveys, even though both involve splitting units into similar groups.
Blocking connects design (Unit 3) to inference (Unit 7) because less variability at the design stage means less worry about inference errors later.
Blocking is an experimental design technique where you group similar experimental units into blocks based on a variable known to affect the response, then randomly assign treatments within each block. It reduces variability so the treatment effect is easier to detect.
No, and this is one of the most common AP Stats mix-ups. Stratified sampling is for surveys (selecting a representative sample from a population), while blocking is for experiments (assigning treatments to reduce variability). If treatments are being assigned, it's blocking.
No. Blocking handles a known confounding variable, but you still must randomly assign treatments within each block to balance out unknown variables. A blocked design without randomization isn't a valid experiment and won't earn FRQ credit.
Blocking removes the variability caused by a known nuisance variable from the error variance. With less background noise, a real treatment effect stands out more clearly, so the test is more likely to detect it. That's the definition of higher power.
Yes. It appears in design-focused multiple-choice questions (like choosing the right method for a field with a moisture gradient) and on FRQs, including 2024 FRQ Q3, where researchers comparing mileage across four car models needed a design accounting for model differences.