What is the difference between random selection and random assignment?

Random selection decides who gets studied and unlocks generalizing to the sampled population. Random assignment decides who gets which treatment and unlocks cause and effect claims. Each one earns its own conclusion, and the SAT tests them separately.

When can a study claim causation?

Only when subjects were randomly assigned to conditions, as in an experiment. If people chose their own behavior, like how much they sleep, the study is observational and supports association language only.

Can results from one city generalize to the whole country?

No. Random selection from one city supports conclusions about that city. Stretching to all adults, all teenagers, or the whole country is the classic wrong answer choice on these questions.

Does a bigger sample fix a biased one?

No. Surveying 10,000 volunteers still measures volunteers. Size shrinks random error, but only random selection deals with bias, and no sample size substitutes for it.

How do I practice a skill that has no calculations?

With repetitions. Satified generates a fresh study description every time: you sort it, selected or not, assigned or not, then match the strongest honest conclusion. It becomes fast pattern recognition, free and without an account.

Data Analysis · Skill 7 of 7

SAT Statistical Claims Practice

These are the digital SAT questions with no algebra at all: a study is described and you choose the strongest conclusion it can honestly support. The wrong answers overreach on purpose, so the skill is knowing exactly what random selection and random assignment each buy you, and refusing everything they do not.

Domain: Data Analysis
About 15% of the test is Data Analysis
Difficulty: easy to hard
Free, no account

Drill statistical claims now → See all Data Analysis skills

The patterns the SAT actually uses

Every one of these questions sorts into a two by two grid: was the sample randomly selected, and were the treatments randomly assigned. Sort first, conclude second.

Pattern 01

Random selection only

Subjects were randomly chosen but nobody assigned their behavior. The result generalizes to the sampled population, and the honest conclusion says association, never cause.

Pattern 02

Random assignment only

Volunteers were randomly split into treatment groups. Cause and effect is on the table, but only for people like those volunteers, not for a whole population.

Pattern 03

Both or neither

Both randomizations together support the strongest claims. Neither supports almost nothing beyond the people actually studied. Place every study in this grid before reading the choices.

Pattern 04

Spot the biased sample

Mall shoppers, website volunteers, one homeroom. Convenience samples poison generalization no matter how large they are, and the SAT plants a big number to tempt you.

One worked example, start to finish

Worked example · medium

Researchers randomly selected 500 adults from one city and recorded nightly sleep and headache frequency. Of the 200 adults who slept fewer than 6 hours a night, 90 reported frequent headaches. Of the 300 adults who slept at least 6 hours, 60 reported frequent headaches. Which conclusion do the results best support?

Compute both rates. Short sleepers: 90 ÷ 200 = 0.45, so 45 percent. Longer sleepers: 60 ÷ 300 = 0.20, so 20 percent. The groups check out: 200 + 300 = 500.
The rates differ sharply, so the data shows an association between short sleep and frequent headaches.
Selection was random and came from one city, so the association generalizes to adults in that city, and no further.
Sleep was observed, not assigned, so nothing here supports a cause and effect claim in either direction.

Answer: An association for adults in this city, not a causal claim

Sleep becomes coffee, exercise, or screen time on regeneration. The design decides the conclusion, every single time.

Where students lose the point

Hearing causation in an observational study. If people chose their own sleep, diet, or screen time, the study can only support that the variables move together, no matter how dramatic the gap.
Generalizing past the sample frame. Randomly selected adults from one city speak for that city's adults. Choices that quietly widen to all adults or all people are wrong on purpose.
Letting sample size impress you. A large convenience sample is still a convenience sample. Ten thousand volunteers are still volunteers, and random beats big every time.
Mixing up the two randoms. Random selection is about who gets studied. Random assignment is about who gets which treatment. Each unlocks a different conclusion, and swapping them flips the answer.

Using Desmos here

Honestly, Desmos does not help here, and that is the defining feature of this skill. There is nothing to graph and barely anything to compute. These are reading questions: who was selected, what was assigned, which population was sampled. Close the calculator and put your energy into the wording of the answer choices, because that is where every one of these questions is decided.

Why drilling here is different

Judgment skills need volume, so Satified generates a fresh study description every time: new populations, new designs, new tempting overreaches, at easy, medium, and hard difficulty. Sorting selection from assignment becomes pattern recognition. Each of the 1,483 questions in the bank has an independently verified answer and explanation.

A new study to judge, every load.

Start this skill free →

Questions students ask

What is the difference between random selection and random assignment?: Random selection decides who gets studied and unlocks generalizing to the sampled population. Random assignment decides who gets which treatment and unlocks cause and effect claims. Each one earns its own conclusion, and the SAT tests them separately.
When can a study claim causation?: Only when subjects were randomly assigned to conditions, as in an experiment. If people chose their own behavior, like how much they sleep, the study is observational and supports association language only.
Can results from one city generalize to the whole country?: No. Random selection from one city supports conclusions about that city. Stretching to all adults, all teenagers, or the whole country is the classic wrong answer choice on these questions.
Does a bigger sample fix a biased one?: No. Surveying 10,000 volunteers still measures volunteers. Size shrinks random error, but only random selection deals with bias, and no sample size substitutes for it.
How do I practice a skill that has no calculations?: With repetitions. Satified generates a fresh study description every time: you sort it, selected or not, assigned or not, then match the strongest honest conclusion. It becomes fast pattern recognition, free and without an account.

Keep going

This caps the Data Analysis domain. If any of the earlier skills felt soft, loop back through them before test day.