When do I use t instead of z for a confidence interval about a mean?

Use the t-distribution whenever the population standard deviation is unknown and you estimate it with the sample standard deviation s. On the AP exam this is nearly always the case for means. You would only use z if were actually given, which is rare.

What are the degrees of freedom for a one-sample t-interval?

For a one-sample t-procedure, df = n - 1, where n is the sample size. This value tells you which t-distribution to use when finding t^*. If a table doesn't list your exact df, round down to the nearest available value for a conservative (slightly wider) interval.

How do I check the Normal condition for small samples?

When n < 30, you cannot rely on the Central Limit Theorem, so you must look at a graph of the sample data—a dotplot, histogram, stemplot, or boxplot. If it shows no strong skew and no outliers, the Normal condition is reasonably met. If it shows strong skew or outliers, the t-interval is not appropriate.

How is interpreting the interval different from interpreting the confidence level?

Interpreting the interval focuses on one specific interval: "We are 95% confident this interval captures the true mean (in context)." Interpreting the confidence level describes the method over many samples: "About 95% of intervals built this way from repeated random samples would contain the true mean." AP questions may ask for either, so know both.

AP-STATS-7.1-7.4

U7.1 One-Sample t-CI for a Mean

Q: How do I check the Normal condition for small samples?

When n < 30, you cannot rely on the Central Limit Theorem, so you must look at a graph of the sample data—a dotplot, histogram, stemplot, or boxplot. If it shows no strong skew and no outliers, the Normal condition is reasonably met. If it shows strong skew or outliers, the t-interval is not appropriate.

Q: How is interpreting the interval different from interpreting the confidence level?

Interpreting the interval focuses on one specific interval: "We are 95% confident this interval captures the true mean (in context)." Interpreting the confidence level describes the method over many samples: "About 95% of intervals built this way from repeated random samples would contain the true mean." AP questions may ask for either, so know both.

Learn to build one-sample t confidence intervals for a population mean when σ is unknown—checking conditions, computing t*, and interpreting results the AP way.

What you'll do in this lesson

A voice-first session with the Crimsora tutor on U7.1 One-Sample t-CI for a Mean, then targeted practice and FRQs — with the tutor adapting to where you get stuck.

Start U7.1 One-Sample t-CI for a Mean free

What this lesson covers

When you want to estimate a population mean

\mu

but you don't know the population standard deviation

\sigma

, you can't use a z-interval. Instead, you estimate

\sigma

with the sample standard deviation

s

and switch to the

t

-distribution, which has heavier tails to account for that extra uncertainty.

This lesson shows you exactly how to construct a one-sample

t

confidence interval for a mean: how to check the required conditions, find the critical value

t^*

using degrees of freedom, compute the margin of error, and write an interpretation that earns full credit. Master this now—every mean-based procedure in Unit 7 builds on these ideas.

Why the t-distribution (not z)?

When

\sigma

is known, the sampling distribution of

\bar{x}

is standardized with

z = \frac{\bar{x}-\mu}{\sigma/\sqrt{n}}

. But in almost every real study,

\sigma

is unknown. We replace it with the sample standard deviation

s

, giving the statistic

t = \frac{\bar{x}-\mu}{s/\sqrt{n}}

.

Because

s

varies from sample to sample, this statistic is more variable than

z

, especially for small samples. The

t

-distribution accounts for that: it is bell-shaped and symmetric like the standard Normal but has heavier tails, so critical values

t^*

are larger than the corresponding

z^*

. This makes intervals wider, correctly reflecting the added uncertainty from estimating

\sigma

.

The shape of a

t

-distribution depends on its degrees of freedom,

df = n - 1

. As

df

increases, the tails thin out and the

t

-distribution approaches the standard Normal.

Feature	z-distribution	t-distribution
Requires	$\sigma$ known	$\sigma$ unknown, use $s$
Shape	Fixed Normal	Depends on $df=n-1$
Tails	Standard	Heavier (wider CIs)
Critical value	$z^*$	$t^*$ (larger)

A common misconception: students think large samples "switch" you to

z

. On the AP exam, whenever

\sigma

is unknown you use

t

, regardless of sample size.

Checking the conditions

Before building any

t

-interval you must verify conditions, and the AP exam awards points specifically for this.

First, the Random condition: the data must come from a random sample or randomized experiment. This lets you generalize to the population and justifies using the sampling distribution.

Second, the Independence (10%) condition: when sampling without replacement, the sample size should be less than 10% of the population,

n \le 0.10N

. This keeps observations approximately independent.

Third, the Normal/Large condition for the sampling distribution of

\bar{x}

. You satisfy it in one of three ways: the population is stated to be approximately Normal; the sample size is large,

n \ge 30

, so the Central Limit Theorem applies; or, for smaller samples, a graph of the sample data (dotplot, histogram, or boxplot) shows no strong skew and no outliers.

When writing conditions, be specific. Don't just say "

n \ge 30

"—state the actual value. For small samples, you must reference the graph of the data, not just claim normality. On free-response questions, failing to check the graph for small

n

is one of the most common ways students lose the condition point.

Constructing the interval

The one-sample

t

-interval for a mean has the familiar form of estimate plus-or-minus margin of error:

\bar{x} \pm t^* \cdot \frac{s}{\sqrt{n}}

Here

\bar{x}

is the sample mean,

s

is the sample standard deviation,

n

is the sample size, and

\frac{s}{\sqrt{n}}

is the standard error of the mean. The critical value

t^*

comes from the

t

-distribution with

df = n-1

and the chosen confidence level.

To find

t^*

, use a

t

-table (find the row for

df

and the column for your confidence level) or technology (invT). For example, a 95% interval with

n=20

uses

df=19

and

t^* \approx 2.093

.

The margin of error

ME = t^* \cdot \frac{s}{\sqrt{n}}

tells you how far the interval extends on each side. Notice three levers: higher confidence increases

t^*

(wider interval), larger

n

shrinks the standard error (narrower interval), and more variable data (larger

s

) widens it.

If a table lacks your exact

df

, the conservative choice is to round down to the next available

df

, which gives a slightly larger

t^*

and a slightly wider (safer) interval. On the calculator, TInterval computes everything directly, but you should still show the formula and values to earn method points on free response.

Interpreting the interval and the confidence level

AP graders distinguish sharply between interpreting an interval and interpreting a confidence level—learn both templates.

To interpret a specific interval, say: "We are 95% confident that the interval from (lower) to (upper) captures the true mean (context) of (population)." Always name the parameter in context and refer to the population, not the sample.

To interpret the confidence level itself, describe the long-run capture rate of the method: "If we took many random samples of this size and built a confidence interval from each, about 95% of those intervals would contain the true population mean."

Avoid these classic errors. Do not say "there is a 95% probability that

\mu

is in this interval"—once computed, the interval either does or doesn't contain

\mu

; the probability language applies to the process, not one interval. Do not talk about 95% of the data or 95% of sample means being in the interval; the interval estimates a mean, not individual values. And never say you are confident about

\bar{x}

—you already know

\bar{x}

; the interval is about the unknown

\mu

.

Exam prompts also ask whether a claimed value is plausible: if that value lies inside the interval, it is a plausible value for

\mu

; if it lies outside, the data provide evidence against it.

Key terms

t-distribution.: A symmetric, bell-shaped distribution with heavier tails than the Normal, used for inference about a mean when $\sigma$ is unknown. Its shape depends on degrees of freedom.
Degrees of freedom (df).: For a one-sample $t$ -procedure, $df = n - 1$ . It determines which $t$ -distribution to use and grows with sample size.
Standard error of the mean.: The estimated standard deviation of $\bar{x}$ , equal to $\frac{s}{\sqrt{n}}$ , used when $\sigma$ is unknown.
Critical value (t*).: The multiplier from the $t$ -distribution, based on confidence level and $df$ , that sets the width of the interval.
Margin of error.: The amount added and subtracted from $\bar{x}$ , equal to $t^* \cdot \frac{s}{\sqrt{n}}$ ; it reflects sampling variability at the chosen confidence level.
Normal/Large condition.: The requirement that the sampling distribution of $\bar{x}$ be approximately Normal, met if the population is Normal, if $n \ge 30$ , or if a graph of the sample shows no strong skew or outliers.
Confidence level.: The long-run proportion of intervals, built by the same method from repeated random samples, that would capture the true parameter.

Worked example

A researcher measures the caffeine content (mg) of a random sample of 15 cups of coffee from a large café chain. The sample mean is

\bar{x} = 118

mg with sample standard deviation

s = 12

mg. A dotplot of the data is roughly symmetric with no outliers. Construct and interpret a 95% confidence interval for the mean caffeine content.

Start by identifying the procedure:

\sigma

is unknown, so use a one-sample

t

-interval for a mean.

Check conditions. Random: the 15 cups are a random sample—satisfied. Independence: it is reasonable that 15 cups is less than 10% of all cups the chain serves—satisfied. Normal/Large:

n = 15 < 30

, but the dotplot is roughly symmetric with no outliers, so the Normal condition is reasonable.

Find the critical value. With

df = n - 1 = 14

and 95% confidence,

t^* \approx 2.145

.

Compute the standard error:

\frac{s}{\sqrt{n}} = \frac{12}{\sqrt{15}} \approx 3.098

.

Margin of error:

ME = t^* \cdot \frac{s}{\sqrt{n}} = 2.145 \times 3.098 \approx 6.65

.

Build the interval:

118 \pm 6.65

, giving approximately

(111.35, 124.65)

mg.

Interpret: We are 95% confident that the interval from about 111.35 mg to 124.65 mg captures the true mean caffeine content of all cups of coffee served by this chain. Because 120 mg lies inside this interval, 120 mg is a plausible value for the true mean.

Practice questions

A 90% confidence interval for a mean is constructed from a sample of size 25 with

\sigma

unknown. Which critical value and degrees of freedom are correct?

$t^*$ with $df = 25$
$z^*$ with no degrees of freedom
$t^*$ with $df = 24$
$t^*$ with $df = 26$

Answer: $t^*$ with $df = 24$

Because

\sigma

is unknown, you use the

t

-distribution, not

z

. Degrees of freedom for a one-sample

t

-procedure equal

n - 1 = 25 - 1 = 24

. The confidence level affects the value of

t^*

but not the

df

A student writes: "There is a 95% probability that the true mean lies between 111 and 125 mg." Explain what is wrong with this interpretation and give a correct one.

Answer: The statement misuses probability for a fixed interval; a correct version describes 95% confidence that the method's interval captures the true mean.

Once the interval

(111, 125)

is computed, the true mean

\mu

is either inside it or not—there is no probability attached to a single fixed interval. The 95% refers to the long-run success rate of the method. A correct interpretation: "We are 95% confident that the interval from 111 to 125 mg captures the true mean caffeine content," and the confidence level means about 95% of intervals from repeated random samples would contain

\mu

A sample of size 8 is taken from a population, and a boxplot of the data shows a strong right skew with one high outlier. Is it appropriate to construct a one-sample t-interval? Explain.

Answer: No. With only $n = 8$ , the Central Limit Theorem does not apply, and the strong skew and outlier violate the Normal/Large condition, so a t-interval is not appropriate.

For small samples (

n < 30

) you rely on the population being approximately Normal, judged from a graph of the data. Strong skew and an outlier indicate the sampling distribution of

\bar{x}

may not be approximately Normal, so the procedure's assumptions fail. You would need a larger sample or a different approach.

FAQ

When do I use t instead of z for a confidence interval about a mean?: Use the $t$ -distribution whenever the population standard deviation $\sigma$ is unknown and you estimate it with the sample standard deviation $s$ . On the AP exam this is nearly always the case for means. You would only use $z$ if $\sigma$ were actually given, which is rare.
What are the degrees of freedom for a one-sample t-interval?: For a one-sample $t$ -procedure, $df = n - 1$ , where $n$ is the sample size. This value tells you which $t$ -distribution to use when finding $t^*$ . If a table doesn't list your exact $df$ , round down to the nearest available value for a conservative (slightly wider) interval.
How do I check the Normal condition for small samples?: When $n < 30$ , you cannot rely on the Central Limit Theorem, so you must look at a graph of the sample data—a dotplot, histogram, stemplot, or boxplot. If it shows no strong skew and no outliers, the Normal condition is reasonably met. If it shows strong skew or outliers, the $t$ -interval is not appropriate.
How is interpreting the interval different from interpreting the confidence level?: Interpreting the interval focuses on one specific interval: "We are 95% confident this interval captures the true mean (in context)." Interpreting the confidence level describes the method over many samples: "About 95% of intervals built this way from repeated random samples would contain the true mean." AP questions may ask for either, so know both.

Learn this with a teacher, not a page

The Crimsora tutor teaches U7.1 One-Sample t-CI for a Mean live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.

Try a live session free