When a poll says '52% support the measure, plus or minus 3 points,' that little 'plus or minus' is the whole point — and the test loves to check whether you actually understand what it means.
A sample of 52% with ±3 points gives a confidence interval of 49% to 55% for the population.
Loading…
Sample size vs. margin of error
Sample size
Effect on margin of error
Why
Small (e.g. 250)
Larger margin
Less representative of population
Large (e.g. 1000)
Smaller margin
More representative; scales with 1/√n
×4 the size
Margin cut in half
√4 = 2
Bigger samples shrink the margin of error.
Loading…
Quick check
A biologist tags 40 fish in a lake and releases them. A week later, she catches a sample of 100 fish and finds that 8 are tagged. Using the capture-recapture method, what is the estimated total number of fish in the lake?
Worked examples
Example 1
Loading…
Example 2
Loading…
Example 3
Loading…
Common pitfalls
Applying the interval to the sample instead of the population
Loading…
Thinking bigger sample = bigger margin of error
Loading…
Forgetting the √n relationship
Loading…
Ignoring whether the sample was random
Loading…
Key takeaways
Loading…
Loading…
Loading…
Loading…
Loading…
Tracks your progress across lessons.
Checkpoint
10 questions on Inference from Sample Statistics and Margin of Error, from our question bank. Score 70% or better across 10 answers to reach Silver and pass this step.
When researchers want to know what a whole population thinks, they can't ask everyone. So they take a sample — a smaller group — and measure that. But a sample is never a perfect copy of the population, so the result comes with a margin of error: a cushion of uncertainty around the sample's number.
Here's the key idea. If a survey says 52% with a margin of error of ±3 percentage points, you build an interval:
Low end: 52% - 3% = 49%
High end: 52% + 3% = 55%
This is called a confidence interval. The correct interpretation is: the true value for the whole population is plausibly between 49% and 55%. That range is your answer to 'what does the population really think?'
Three things the test checks over and over:
1. The interval is about the POPULATION, not the sample. We already know exactly what the sample said (52%). The margin of error tells us how far the population's true value might be from that sample number. Wrong answers often say 'between 49% and 55% of the sample' — that's a trap.
2. Bigger sample → smaller margin of error. The more people you ask, the closer your sample gets to the truth, so the cushion shrinks. Specifically, the margin of error is inversely proportional to the square root of the sample size (√n). To cut the margin in half, you must make the sample 4 times bigger (because √4 = 2).
3. Random sampling matters. The margin of error formula only works if the sample was chosen randomly from the population. If the survey only asked people leaving a gym, you can't generalize to everyone — no margin of error fixes a biased sample.
You almost never have to calculate a margin of error on the test. Instead you interpret it (build the interval, describe what it means) or reason about it (what happens if the sample gets bigger or smaller). Get those two skills down and this topic is easy points.
A poll of 600 registered voters found that 47% favor a new transportation bill, with a margin of error of ±4 percentage points. Based on this poll, the proportion of ALL registered voters who favor the bill is plausibly between what two values?
A survey of 250 randomly selected students found that 70% attend at least one school event per year, with a margin of error of ±6 percentage points at a 95% confidence level. If the researchers had instead surveyed 1,000 randomly selected students, what would most likely happen to the margin of error?
A researcher reports that 38% of a random sample of 900 commuters use public transit, with a margin of error of ±3 percentage points. Which statement is the BEST interpretation of this result?
We already know the sample's exact percentage — the margin of error is about how far the whole population's true value might be. Any answer that says 'X% of the sample' is wrong.
It feels like 'more people = more spread,' but it's the opposite. A larger sample is more representative, so the margin of error shrinks. More data = more precision.
To halve the margin of error you need 4× the sample, not 2×, because margin of error scales with 1/√n. Doubling the sample only shrinks it by a factor of √2 ≈ 1.4.
Margin of error only lets you generalize to the population if the sample was randomly selected. A convenience sample (e.g. only people at a gym) can't be fixed by any margin of error.
Confidence interval = sample percent ± margin of error (subtract and add).
The interval describes the true POPULATION value, not the sample.
Larger sample size → smaller margin of error (inversely proportional to √n).
To halve the margin of error, multiply the sample size by 4.