Digital SAT Math: Sampling, Margin of Error and Cause vs Correlation

Audience/search intent: SAT students seeking to interpret statistical claims and avoid common inference errors, rather than solve algebraic equations or calculate descriptive statistics.

A digital SAT statistics question may contain almost no difficult arithmetic. The challenge is deciding what a survey or experiment actually tells us. Can a result from 240 students describe a school of 1,200? Does a poll showing 62% support mean that exactly 62% of the population agrees? If two groups receive different teaching methods, can we say that one method caused the difference?

These questions belong to the SAT Math Problem-Solving and Data Analysis domain. College Board explicitly lists inference from sample statistics and margin of error and evaluating statistical claims: observational studies and experiments. The SAT tests interpretation, not just computation.

1. Start by identifying the population and sample

The population is the entire group a claim concerns. A sample is the smaller group actually observed. The statistic calculated from a sample is used to estimate a population value, but it is not guaranteed to match that value exactly.

Imagine a school with 1,200 students. Researchers randomly select 240 students, and 138 say they prefer a later start time.

The sample proportion is:

$$\frac{138}{240}=0.575=57.5\%.$$

If the sample represents the school reasonably well, an estimate of the number of students who prefer a later start is:

$$0.575\times1200=690.$$

Answer: About 690 students, estimated—not a confirmed count. The school did not ask all 1,200 students. A multiple-choice option that says “exactly 690 students prefer a later start” makes a stronger claim than the evidence supports.

A sample selected randomly from the relevant population is generally more defensible than one gathered by asking only the students who happen to be in the cafeteria at lunch. Random selection does not eliminate uncertainty, but it helps reduce selection bias.

2. Read a margin of error as a range, not an adjustment to the sample size

Suppose a poll reports that 62% of a population supports a proposal, with a margin of error of 4 percentage points.

The reported interval is:

$$62\%-4\%=58\%,\qquad62\%+4\%=66\%.$$

Interpretation: The estimate is 62%, with an indicated range from 58% to 66%. The margin of error concerns uncertainty in the population estimate. It does not mean that 4% of survey responses were wrong, or that the number of people surveyed should be increased by 4%.

The distinction between percentage points and percent change matters. Moving from 62% to 66% is an increase of 4 percentage points; it is not a 4% relative increase. A relative 4% increase from 62% would instead be 62% × 1.04 = 64.48%.

College Board's current math specifications include interpreting margins of error. The examples here focus on interpreting a reported margin, rather than deriving a statistical formula. Without information about the poll's method or confidence level, avoid asserting a particular probability that the true value falls within the stated interval.

3. Why larger random samples usually help

Two otherwise comparable, well-designed random polls ask the same question. Poll A surveys 200 people; Poll B surveys 800.

Which poll would generally have the smaller sampling margin of error? Poll B. A larger random sample usually reduces sampling variability, provided the sampling design and other relevant conditions are comparable.

But sample size is not a cure for a biased method. Surveying 8,000 volunteers from one fan club can be less representative of a city than randomly surveying a much smaller number of city residents. The key question is not only “How many?” but also “How were they chosen?”

4. Random sampling and random assignment do different jobs

These two phrases sound similar but answer different questions.

Random sampling concerns who enters the study. It helps researchers make a defensible generalisation from a sample to the population it came from.

Random assignment concerns which treatment participants receive. In a well-controlled experiment, it helps support a causal interpretation because it tends to balance other influences between groups.

Consider 80 students who volunteer for a study of two revision methods. Researchers randomly assign 40 to Method A and 40 to Method B. After the study, 30 students in Group A and 22 in Group B meet a target score.

Method A success rate: 30 ÷ 40 = 75%.

Method B success rate: 22 ÷ 40 = 55%.

Difference: 20 percentage points.

Because the study used random assignment, the observed difference can provide evidence of a causal effect of the revision method within the study, assuming the experiment was properly conducted. But the 80 students volunteered; they were not randomly sampled from all teenagers. We cannot automatically generalise the result to every SAT student.

Now change the design. Suppose the students chose their own revision method. Even if Method A still scored higher, the difference could partly reflect motivation, prior attainment, or another factor. That is an observational comparison, not the same causal evidence as random assignment.

5. Five SAT traps to avoid

• Sample versus population: A sample result estimates a population quantity; it does not count every individual.

• Margin of error: Add and subtract percentage points from a reported percentage, not from the sample size.

• Larger sample: Bigger often reduces random sampling error, but cannot fix systematic selection bias.

• Correlation versus causation: Two variables moving together does not prove one caused the other.

• Random selection versus random assignment: The first supports generalisation; the second strengthens causal inference.

When reading an SAT question, underline its actual claim: estimate, all students, caused, likely, or randomly selected. These words often determine which answer is justified.

6. Four original SAT-style practice questions

Question 1 — Estimate a population count. A random sample of 150 students finds that 96 bring lunch from home. A district has 800 students. Estimate how many bring lunch from home.

Question 2 — Interpret an interval. A survey reports 47% support for a new rule, with a margin of error of 3 percentage points. What interval is indicated?

Question 3 — Identify the design. Researchers randomly select 120 pupils from a school and ask how much homework they do. They find that pupils who report more homework also report higher grades. Does this establish that homework causes higher grades? Explain.

Question 4 — Decide what can be generalised. A group of 60 volunteers is randomly assigned to two study apps. The group using App A performs better. Which conclusion is most defensible: (a) App A definitely benefits all students; (b) the experiment offers evidence that App A affected results among the study participants; or (c) random assignment proves the volunteers represent every student?

Answers and reasoning

1. 96 ÷ 150 = 0.64; 0.64 × 800 = 512 students estimated.

2. 47 − 3 = 44 and 47 + 3 = 50, so 44% to 50%.

3. No. The study observes an association but does not randomly assign homework amounts. Prior ability, family support and other variables could contribute to the relationship.

4. (b). Random assignment supports a causal comparison within a properly conducted experiment. Volunteer recruitment does not by itself justify generalisation to all students.

7. A focused 15-minute revision routine

Spend five minutes labelling the population, sample, statistic and claimed conclusion in three short scenarios. Spend five minutes translating reported percentages and margins of error into intervals. Finish with five minutes deciding whether a study supports a population generalisation, a causal conclusion, both, or neither.

The most useful habit is to justify the scope of a conclusion: Who was studied, how were they chosen, and what was actually changed or measured? That reasoning is often more valuable than pressing buttons on a calculator.

Official references and fact-check notes — checked 28 October 2026

• College Board: Problem-Solving and Data Analysis — confirms inference from sample statistics, margin of error, and evaluation of observational studies and experiments.

• College Board: Math Specifications — confirms basic study design and interpretation of margin of error within SAT Math.

• College Board: SAT Math Section Overview — confirms the four domains and contextual nature of some questions.

Editorial accuracy note: All numerical examples and practice items above are original and invented for instruction; they are not official SAT questions. The 62% ± 4 percentage points example is an illustrative reported margin, not a computed statistical confidence interval. No specific confidence level is claimed.

---