In AP Statistics, an experiment is a study where researchers deliberately impose treatments (levels of an explanatory variable) on experimental units, using random assignment to balance confounding variables. It's the only study type that lets you conclude cause and effect.
An experiment is a study where researchers actually do something to the participants. They impose different conditions, called treatments, on experimental units and then measure a response variable. The treatments are the levels (or combinations of levels) of one or more explanatory variables, also called factors. This is the dividing line in the CED's definition of study types (3.2.A): in an observational study, researchers just watch and record; in an experiment, they intervene.
A well-designed experiment has four ingredients (3.5.B): comparison of at least two treatment groups (one may be a control group), random assignment of treatments to units, replication (more than one unit per group), and control of potential confounding variables. Random assignment is the magic ingredient. It tends to balance out the effects of variables you can't control, so when the groups end up with different responses, you can attribute the difference to the treatments themselves. That's why experiments, and only experiments, support cause-and-effect conclusions.
Experiments anchor Unit 3 (Collecting Data), specifically Topic 3.2 (identifying study types, AP Stats 3.2.A) and Topic 3.5 (experimental design, AP Stats 3.5.A, 3.5.B, and 3.5.C). The CED is blunt about the payoff: you cannot determine causal relationships from observational data (3.2.B), so 'experiment vs. observational study' decides what conclusions are even legal. The concept then resurfaces in Unit 7, where the conditions for a two-sample t-interval (7.6.B) explicitly accept 'a randomized experiment' as a way to satisfy independence. Design in Unit 3 and inference in Unit 7 are two ends of the same pipeline. How you collect the data determines what your confidence interval is allowed to say about it.
Keep studying AP Statistics Unit 9QyrTKQ1AuLyQwom
Random Assignment (Unit 3)
Random assignment is what makes an experiment an experiment. It spreads uncontrolled variables roughly evenly across treatment groups, so a difference in responses can be pinned on the treatment. Don't confuse it with random sampling, which is about who gets into the study, not which treatment they get.
Control Group (Unit 3)
A control group gives you a baseline for comparison, either receiving no treatment or a placebo. The CED says a well-designed experiment needs at least two groups to compare, and a control group is the classic second group.
Causal Relationships (Unit 3)
This is the whole point. Experiments earn cause-and-effect conclusions because random assignment handles confounding. The coffee-and-longevity practice question is a classic setup: an observational study can show association, but only a randomized experiment could conclude that coffee causes longer life.
Confidence Intervals for a Difference of Two Means (Unit 7)
When you compare two treatment groups on a quantitative response, the two-sample t-interval from Topic 7.6 is how you formalize the comparison. The independence condition (7.6.B) is satisfied by either two independent random samples or a randomized experiment, so your Unit 3 design literally checks a Unit 7 box.
Experiments show up two ways. On multiple choice, you'll classify a study as an experiment or an observational study (records from 500 heart-attack patients, with no treatment imposed, is observational) and judge what conclusion is justified. The trap answer always lets an observational study claim causation. On FRQs, experimental design is a near-annual fixture. The 2019 FRQ asked for a completely randomized design assigning fungus concentrations to trees, the 2018 FRQ involved ACL surgery recovery treatments, the 2021 FRQ had you design a walking-and-cholesterol experiment, and the 2017 FRQ explored random assignment of four people to treatment and control groups. You need to describe a concrete randomization method (random number generator, drawing chips), name the units, treatments, and response variable, and explain why random assignment permits a causal conclusion. Vague answers like 'randomly split them up' lose points; graders want a mechanism.
The difference is one verb: impose. In an experiment, researchers impose treatments on units; in an observational study, they measure variables without intervening. The consequence is huge. Experiments with random assignment can establish cause and effect, while observational studies can only show association because confounding variables remain uncontrolled. Pulling medical records to compare aspirin-takers with non-takers is observational, even though it compares two groups, because nobody assigned anyone to take aspirin.
An experiment imposes treatments on experimental units, while an observational study only measures what's already happening.
A well-designed experiment needs four things: comparison of at least two groups, random assignment of treatments, replication, and control of confounding variables where appropriate.
Random assignment balances confounding variables across groups, which is exactly why experiments can justify cause-and-effect conclusions and observational studies cannot.
The treatments are the levels of the explanatory variable (factor), and the response variable is what you measure after treatments are administered.
A randomized experiment satisfies the independence condition for two-sample inference in Unit 7, so good design in Unit 3 directly enables stronger conclusions later.
On FRQs, describe a specific randomization mechanism, like a random number generator or drawing chips, rather than just saying 'assign randomly.'
An experiment is a study where researchers impose treatments (levels of an explanatory variable) on experimental units and measure a response variable. With random assignment, comparison groups, and replication, it's the only study type that can establish cause and effect.
No. The CED states it directly: you cannot determine causal relationships from observational data because no treatments are imposed and confounding variables aren't controlled. Only a randomized experiment supports causal conclusions.
Random sampling is how you select subjects from the population, and it lets you generalize results to that population. Random assignment is how you split subjects into treatment groups within an experiment, and it lets you conclude causation. An experiment can have one without the other.
Not necessarily. The CED requires comparison of at least two treatment groups, one of which could be a control group. Comparing a low-carb diet group to a high-protein diet group is a valid experiment with no control group at all.
Yes, heavily. Designing or critiquing an experiment appeared on released FRQs in 2017, 2018, 2019, and 2021, including designs assigning fungus treatments to trees and a walking-and-cholesterol study. Expect to name units, treatments, the response variable, and a specific randomization method.