On this concept: Formula

Not sure which lesson to do first? Find your level and get a plan →

Statistics · Grade 9-12

Confidence Interval

A confidence interval is a range of values, calculated from sample data, constructed so that the procedure captures the true population parameter a specified percentage of the time (e.

estimate±margin of error

Quick Answer

A confidence interval is a range of values, calculated from sample data, constructed so that the procedure captures the true population parameter a specified percentage of the time (e.g., 95%). Use it when you estimate a population mean or proportion from a sample and want to report how far off the estimate could be. You build it as estimate ± critical value × standard error. A poll result of 52% with a 95% interval of 49% to 55% says that every value in that range is a plausible level of support among all voters.

Why This Matters

A single estimate hides its own uncertainty. "3.1 hours of screen time a day" sounds exact, but from 100 students the true school average could easily be 2.8 or 3.4. The interval makes that visible, lets you compare two groups honestly, and answers questions like "is 3 hours plausible?". Misread the 95% as the share of students inside the range, or as the chance this one interval is right, and you will claim more than the sample shows.

Intuitive Explanation

Picture a school where students' true average homework time is μ=2.0 hours a night, a number nobody knows. Individual students vary with σ=1.0 hour, so the mean of a random sample of 100 has standard error 1.0/100=0.1 hour. One research team gets xˉ=2.1 and reports 2.1±1.96(0.1), about 1.90 to 2.30 hours: it contains 2.0. Another team draws xˉ=2.25 and reports 2.05 to 2.45: it misses.

Why does the method work 95% of the time? In 95% of samples, xˉ lands within 1.96 standard errors (here 0.196 hours) of μ, and whenever xˉ is within 0.196 of μ, μ is within 0.196 of xˉ, so the interval catches it. The 95% is a property of the recipe, not of any one result. Once a team has its interval, μ is either inside or not. Two dials set the width: higher confidence widens it (99% uses 2.576 instead of 1.96), and a larger sample narrows it, with four times the data halving the width.

A range that holds about 95% of individual students looks similar but is far wider: roughly 2.0±2(1.0), from 0 to 4 hours. That range describes how students differ, and it stays wide however many students you survey. A confidence interval is about one fixed number, the population mean, so it keeps shrinking as n grows. If a range would tighten with more data, it is a confidence interval.

Think of it like this. Ring toss with a peg you cannot see: the peg (the true value) never moves, and each sample throws a ring (the interval) centered on its own estimate. A 95% method rings the peg on about 95 throws out of 100. Once a ring has landed, it either circles the peg or it doesn't; the 95% describes the thrower, not that throw.

Core idea. A sample estimate usually lands within about two standard errors of the truth, so the truth usually lies within two standard errors of the estimate. The confidence level is the long-run hit rate of that procedure, not a probability about any one interval.


When to Use

Build a confidence interval when a question asks for a range of plausible values for a population mean, proportion or difference, or pairs an estimate with a confidence level such as 90%, 95% or 99%. You need the sample statistic, its standard error and a critical value. If only the ± half-width is wanted, that is the margin of error; if the question asks whether one claimed value can be rejected, run a hypothesis test. A two-sided test at α=0.05 rejects roughly the same values that fall outside a 95% interval.

Tip. Ask: Is the goal to report a range of plausible values for an unknown population mean or proportion, built from one sample?

How to Recognize It

Before building or reading an interval, check what unknown number it is meant to capture and how sure the method is.

  1. Is the unknown a population parameter, such as the mean sleep time of all students at a school?

    Yes, a confidence interval estimates it. If the range is meant to hold most individual values, that is a statement about data spread, like the Empirical Rule.

  2. Is a confidence level, such as 90% or 95%, stated or needed?

    Yes fits. A bare spread with no confidence level attached is the Standard Error.

  3. Does the answer need two endpoints, like 6.9 to 7.5 hours?

    Yes means a confidence interval. If only the plus-or-minus amount is asked for, the answer is the Margin of Error.

  4. Is there a specific claimed value to accept or reject, like a mean of 8 hours?

    Then the main job is Hypothesis Testing, though an interval computed from the same data can support that decision.

Confidence Interval vs Margin of Error vs Empirical Rule vs Hypothesis Testing

All four involve a 95% or a plus-or-minus, so they get swapped. Ask whether the range is for a parameter, is only its half-width, holds individual data values, or whether the job is a decision about one claimed value.

Confidence Interval

Meaning
A confidence interval is a range built from sample data by a method that captures the true parameter in a stated share of samples.
Key test
Two endpoints around an estimate of a parameter, with a confidence level.
Formula
xˉ±t∗sn
Example
Mean sleep of all the school's students: 6.9 to 7.5 hours, with 95% confidence.

Margin of Error

Meaning
The margin of error is the plus-or-minus half-width added to and subtracted from the estimate.
Key test
Just the plus-or-minus amount, not the endpoints.
Formula
E=t∗sn
Example
The ±0.3 in 7.2±0.3 hours.

Empirical Rule

Meaning
The empirical rule says about 95% of individual values in a normal distribution lie within 2 standard deviations of the mean.
Key test
A range for single values, much wider than an interval for the mean.
Formula
μ±2σ
Example
About 95% of single students sleep between 4.2 and 10.2 hours, if sleep is roughly normal.

Hypothesis Testing

Meaning
Hypothesis testing checks whether sample data give strong enough evidence against one claimed value of a parameter.
Key test
Ends in a decision about a claim, not a range.
Example
Testing whether mean sleep differs from the recommended 8 hours.

Formula & Notation

estimate±margin of error
A (1−α)100% confidence interval for μ is xˉ±zα/2⋅σn. When σ is unknown, use xˉ±tα/2,n−1⋅sn.

How to read it: CI is the confidence interval. zα/2 is the critical z-value (1.96 for 95%). α is the significance level. The margin of error is E=zα/2⋅SE.

Worked Examples

Example 1 — 95% interval for a proportion

Easy

In a random sample of 400 students at a large school, 240 say they prefer online homework. Build a 95% confidence interval for the proportion of all students at the school who prefer it.
Show the steps and the answer
  1. Compute the sample proportion: p^=240/400=0.60.

    The interval is centered on the estimate from the sample.

  2. Check the counts: np^=240 and n(1−p^)=160, both at least 10.

    With this many successes and failures, the sampling distribution of p^ is close to normal, so 1.96 is the right multiplier.

  3. Find the standard error: SE=0.6×0.4/400=0.0006≈0.0245.

    This is how much p^ typically varies between samples of 400.

  4. Build the interval: 0.60±1.96(0.0245)=0.60±0.048, from 0.552 to 0.648.

    1.96 standard errors on each side capture the middle 95% of the normal curve.

Answer: From about 0.552 to 0.648: we are 95% confident that between 55.2% and 64.8% of all students at the school prefer online homework.

Takeaway: Estimate, standard error, multiplier, then interpret in context: the interval is a statement about the whole school, not about the 400 students asked.

Example 2 — An interval for the mean is not a range for one apple

Standard

A random sample of 100 apples from an orchard has mean weight 150 g and standard deviation 20 g. (a) Find a 95% confidence interval for the mean weight of all the orchard's apples. (b) A shopper wants to know what one apple she picks is likely to weigh. Does the interval from (a) answer her?
Show the steps and the answer
  1. Compute SE=20/100=2 g and the margin 1.96×2=3.92 g.

    An interval for a mean uses the spread of the sample mean, not of single apples.

  2. Build the interval: 150±3.92, from 146.1 g to 153.9 g.

    This is the range of plausible values for the orchard's average apple weight.

  3. Read part (b): it asks about one apple, an individual.

    Single apples vary with the standard deviation of 20 g, ten times the standard error, so the interval from (a) is far too narrow for them.

  4. Give a range for individual apples: about 150±2(20), from 110 g to 190 g, if weights are roughly bell-shaped.

    About 95% of individual values lie within 2 standard deviations of the mean, and this range would not shrink if more apples were sampled.

Answer: (a) 146.1 g to 153.9 g for the mean. (b) No: a single apple typically weighs somewhere between about 110 g and 190 g.

Takeaway: A confidence interval pins down a parameter and narrows as n grows; a range for individuals depends on the standard deviation and stays wide.

Example 3 — Find the mistake: what the 95% refers to

Application

A random sample of 64 commuters has mean commute 32 minutes and standard deviation 12 minutes. The 95% confidence interval is 29.1 to 34.9 minutes. Kai writes: "There is a 95% probability that the true mean commute is between 29.1 and 34.9 minutes." Find the mistake and correct it.
Show the steps and the answer
  1. Check the interval: SE=12/64=1.5, margin 1.96×1.5=2.94, so 32±2.94 gives 29.06 to 34.94.

    The arithmetic is right; the problem is in Kai's sentence.

  2. Ask what is random here: the true mean commute is one fixed number.

    It is either between 29.1 and 34.9 or it is not, so no 95% chance is left over for this particular interval.

  3. Locate the 95%: it is the success rate of the method across repeated samples.

    If many groups each sampled 64 commuters and built an interval this way, about 95% of their intervals would contain the true mean.

  4. Rewrite Kai's sentence: "We are 95% confident the true mean is between 29.1 and 34.9 minutes, because 95% of intervals built this way capture it."

    The confidence is placed in the method, which is all a single sample can support.

Answer: We are 95% confident that the mean commute of all commuters is between 29.1 and 34.9 minutes: intervals built this way capture the true mean in about 95% of samples.

Takeaway: The confidence level describes the procedure, not one computed interval; once the data are in, the true value is either inside or outside.

Where You'll See It

  • Election poll reported as a range

    Instead of saying a candidate has exactly 52% support, a poll reports '52% with a 95% confidence interval of 49% to 55%,' acknowledging the result came from a sample, not every voter.

  • Estimating average screen time at school

    Surveying 100 students gives a mean of 3.1 hours; a 95% confidence interval of 2.7 to 3.5 hours communicates that the true schoolwide average is plausibly anywhere in that band.

  • A drug trial's benefit as a range

    A blood-pressure drug trial reports an average drop of 8 mmHg with a 95% confidence interval of 5 to 11 mmHg. Doctors read it as: the true average benefit could plausibly be as small as 5 or as large as 11, and because 0 lies outside the range, "no effect" is not a plausible value.

Common Mistakes

  1. The slip: A 95% confidence interval contains 95% of the individual data values.

    Why it's tempting: The '95%' feels like it describes how much data is inside the range.

    Instead: A 95% CI describes a procedure: 95% of intervals built this way over repeated sampling capture the true parameter — it is about the parameter, not about where individual data points fall.

  2. The slip: There is a 95% probability the true parameter is inside this one specific interval I computed.

    Why it's tempting: After computing one interval, it is tempting to attach a probability to that single interval.

    Instead: For a single computed interval the parameter is either in it or not; the 95% refers to the long-run success rate of the method, not the probability for one particular interval.

  3. The slip: A 99% confidence interval is narrower than a 95% one, because it is more confident.

    Why it's tempting: More confidence sounds like more precision, and a tighter range feels like a surer answer.

    Instead: To capture μ more often the interval must be wider: 99% uses 2.576 standard errors instead of 1.96. With xˉ=2.1 and SE=0.1 hour, the 95% interval is about 1.90 to 2.30 hours and the 99% interval is about 1.84 to 2.36 hours.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. A 95% confidence interval for the mean commute time is 22.6 to 27.4 minutes. Find the sample mean and the margin of error.

    Hint: The estimate sits in the middle; the margin is half the width.

  2. A random sample of 500 voters finds 46% support a new park, with a 95% margin of error of 4.4 percentage points. Write the confidence interval and say what it means.

    Hint: Subtract and add the margin to the estimate.

  3. A random sample of 64 bags of chips has mean weight 28.4 g, and the population standard deviation is 1.6 g. Build a 95% confidence interval for the mean weight using z∗=1.96.

    Hint: Find the standard error first, then multiply it by 1.96.

  4. A 95% confidence interval for the proportion of teens who sleep 8 or more hours is 0.31 to 0.39. An article claims half of teens sleep that much. Is 0.50 plausible?

    Hint: Check whether 0.50 is inside the interval.

  5. A sample of 40 phones has mean battery life 11.2 hours and standard deviation 1.9 hours. A manager wants a range of believable values for the mean battery life of all phones of this model. Should you report the standard error or a confidence interval?

    Hint: Does the manager want a single spread number or a range of values?

  6. A student writes: "Our 95% confidence interval for the mean height of students is 160 cm to 170 cm, so 95% of students are between 160 and 170 cm tall." Correct the interpretation.

    Hint: Is the interval about individual students or about the average?

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

Why not use a 99% interval every time to be safer?

Because more confidence costs width. A 99% interval uses a critical value of about 2.58 instead of 1.96, so it is about a third wider. An interval wide enough to be almost certain, like 'average sleep is between 0 and 24 hours', tells you nothing useful. Researchers pick a level that balances being right with being precise, and 95% is the common compromise.

If two groups' 95% intervals overlap, does that mean there is no real difference?

Not necessarily. Two intervals from separate groups can overlap a little while the difference between the groups is still statistically significant, because the uncertainty in a difference is smaller than the two margins added together. If the intervals do not overlap at all, the difference is significant. For a proper check, build one interval for the difference itself and see whether it contains 0.

Why do intervals for a mean use a t value instead of 1.96?

Because the population's standard deviation is almost never known, so you estimate it with ss, and that estimate adds extra uncertainty, especially in small samples. The t distribution accounts for it with heavier tails: for 95% confidence the critical value is 2.26 with 10 observations and 2.04 with 31, compared with 1.96 for z. With large samples the two nearly match.

Learning Path

Confidence Interval

You are here

You need Standard Error (how much an estimate wobbles) and Sampling Distribution (why about 95% of estimates land within 1.96 standard errors of the truth). A confidence interval turns those into a range you can report: estimate ±z∗⋅SE. Next, Margin of Error studies the half-width on its own, including how sample size controls it, and Hypothesis Testing asks whether one particular value is believable.

See Also