What Is a t Distribution? — AP Statistics Study Guide

For: AP Statistics candidates sitting AP Statistics.

Covers: Definition of the t distribution, degrees of freedom calculation, shape properties, comparison to the standard normal (z) distribution, conditions for using t procedures for inference about means, and calculation of t test statistics.

You should already know: Normal distribution properties, sampling distributions of the sample mean, how to calculate standard error.

A note on the practice questions: All worked questions in the "Practice Questions" section below are original problems written by us in the AP Statistics style for educational use. They are not reproductions of past College Board / Cambridge / IB papers and may differ in wording, numerical values, or context. Use them to practise the technique; cross-check with official mark schemes for grading conventions.

1. What Is a t Distribution?

The t distribution (also called Student's t-distribution) is a bell-shaped, symmetric probability distribution that we use for inference about population means when the population standard deviation $\sigma$ is unknown, which is almost always the case in real-world AP Statistics problems. This topic is core to Unit 7: Inference for Quantitative Data: Means, which accounts for 12-15% of the total AP Exam weight, and it appears in both multiple-choice (MCQ) and free-response (FRQ) sections, usually as a foundational step for confidence intervals or hypothesis tests for means. The distribution was developed by William Gosset under the pseudonym "Student" when he worked at Guinness, so you may sometimes see it called Student's t-distribution. Unlike the fixed standard normal (z) distribution, the t distribution changes shape based on a parameter called degrees of freedom, which is tied to the sample size. Every inference procedure for a population mean when $\sigma$ is unknown relies on the t distribution, so this is the foundational concept for all the t-confidence intervals and t-tests you will do in this unit.

2. Degrees of Freedom for the t Distribution

Degrees of freedom (often abbreviated df) is the parameter that controls the shape of the t distribution. For all one-sample inference procedures for a population mean, degrees of freedom are calculated as: $$df=n-1$$ where $n$ is the sample size. The reason we subtract 1 from $n$ comes from how we calculate the sample standard deviation $s$: to find $s$, we first use the sample mean $\bar{x}$ as an estimate of the unknown population mean $\mu$. Since the sum of deviations from $\bar{x}$ is always fixed at zero, only $n-1$ of these deviations are free to vary—once you know $n-1$ deviations, the last one is determined automatically. This gives us $n-1$ independent pieces of information, hence $n-1$ degrees of freedom. For two-sample t procedures, df is calculated differently (either the conservative minimum of $n_1-1$ and $n_2-1$, or the Welch-Satterthwaite approximation), but one-sample df is always $n-1$. As df increases, the t distribution becomes narrower and closer to the standard normal distribution, because larger samples give more accurate estimates of $\sigma$, reducing extra uncertainty.

Worked Example

A marine biologist collects a random sample of 18 sea turtle hatchlings to estimate the mean weight of hatchlings at a nesting beach. The population standard deviation of hatchling weights is unknown. What degrees of freedom should she use for a t procedure to estimate the population mean? How would df change if she adds 6 more randomly sampled hatchlings to her data set?

1. This is a one-sample inference problem with unknown $\sigma$, so we use the one-sample df formula $df=n-1$.
2. Original sample size $n=18$, so $df=18-1=17$.
3. After adding 6 more hatchlings, the new sample size is $n=18+6=24$.
4. New degrees of freedom is $df=24-1=23$. Increasing the sample size increases df, which reduces the spread of the t distribution.

Exam tip: Always write your degrees of freedom explicitly on FRQs; AP graders require it to award full credit for inference procedures.

3. Shape Properties: t Distribution vs. Standard Normal (z) Distribution

Both the t distribution and standard normal (z) distribution are unimodal, symmetric, and centered at 0, but they differ in spread. The t distribution has heavier tails than the z distribution for any finite degrees of freedom. This extra spread in the tails comes from the additional uncertainty we get when we estimate $\sigma$ with $s$: unlike $\sigma$, which is a fixed population parameter, $s$ varies from sample to sample, so we have more overall uncertainty in our estimate of $\mu$.

As degrees of freedom increase, the t distribution converges to the z distribution. When df approaches infinity, $s$ becomes almost identical to $\sigma$, so the extra uncertainty disappears, and t equals z exactly. Even for df = 30, the t distribution is very close to z, but it is never exactly the same for finite sample sizes. This difference in tails means that for any given confidence level, the t critical value $t^{\ast }$ is always larger than the corresponding z critical value $z^{\ast }$, leading to wider intervals that account for extra uncertainty.

Worked Example

Compare the 95% confidence critical values for a sample size of $n=6$ and a very large sample size approaching infinite df. What does the difference between these values tell us about uncertainty for small vs large samples?

1. For a very large sample (infinite df), the t distribution converges to the standard z distribution, so the 95% critical value is $z^{\ast }=1.96$.
2. For $n=6$, degrees of freedom $df=6-1=5$. Looking up 95% confidence for $df=5$ in a t-table gives a critical value of $t^{\ast }=2.571$.
3. The difference between the two critical values is $2.571-1.96=0.611$, meaning the t critical value is 31% larger than the z critical value for this small sample.
4. This larger critical value reflects the higher uncertainty from estimating $\sigma$ with $s$ when the sample size is small: we need a wider interval to capture the true population mean 95% of the time.

Exam tip: If you forget whether t has fatter tails than z, remember: smaller sample = more uncertainty = more spread = fatter tails. This logic works for any MCQ question asking to compare shapes.

4. When To Use the t Distribution

We use the t distribution for inference about a population mean (or difference in two population means) any time the population standard deviation $\sigma$ is unknown, which is almost always true in real-world and AP exam problems. Two core conditions must be satisfied to use t procedures (and hence the t distribution):

1. Random: The data comes from a random sample from the population of interest, or a randomized comparative experiment.
2. Normal/Large Sample: The sampling distribution of $\bar{x}$ is approximately normal. This is true if either the sample size is large ($n\ge 30$, by the Central Limit Theorem), or the population distribution is approximately normal. For small samples ($n<30$), we check the sample data for strong skewness or extreme outliers; if none exist, we can assume the population is approximately normal.

If $\sigma$ is known (a very rare case, almost only used for introductory teaching), we can use the z distribution for inference about means, but this almost never appears on the AP exam for means inference.

Worked Example

A coffee shop owner wants to test whether the mean wait time for a custom drink order is more than 2 minutes. He collects a random sample of 14 order wait times, plots the data, and finds no outliers or strong skewness. The population standard deviation of wait times is not known. Should he use the t distribution for this hypothesis test? Justify your answer.

1. Check the first condition: The sample is random, which is given, so the random condition is satisfied.
2. Check the second condition: The sample size $n=14<30$, but there are no outliers or strong skewness in the sample, so it is reasonable to assume the population distribution of wait times is approximately normal.
3. The key requirement for using the t distribution is an unknown population standard deviation $\sigma$, which is satisfied here.
4. Conclusion: Yes, the t distribution is appropriate for this hypothesis test.

Exam tip: On AP FRQs, you always need to justify using the t distribution by mentioning two things: $\sigma$ is unknown, AND the random and normality conditions are met. Don't just stop at "we use t because $\sigma$ is unknown."

5. Common Pitfalls (and how to avoid them)

• Wrong move: Using the z distribution instead of the t distribution for inference about a mean when $\sigma$ is unknown. Why: Students confuse the rare case of known $\sigma$ with the common real-world case of unknown $\sigma$ that appears on almost all AP exam problems. Correct move: Always use the t distribution for means inference unless the problem explicitly gives you the population standard deviation $\sigma$.
• Wrong move: Calculating degrees of freedom as $df=n$ instead of $df=n-1$ for one-sample t procedures. Why: Students forget that we lose one degree of freedom when we estimate $\mu$ with $\bar{x}$ to calculate $s$. Correct move: Memorize that for one-sample t inference, $df=n-1$ always, and write this explicitly on every problem.
• Wrong move: Claiming the t distribution is skewed or has thinner tails than the z distribution. Why: Students mix up t distribution properties with the chi-square distribution (used for inference on variance/GOF), which is right-skewed. Correct move: Remember t is symmetric, centered at 0, and always has fatter tails than z for any finite degrees of freedom.
• Wrong move: Using the t distribution for inference about a population proportion. Why: Students confuse inference for quantitative means (which uses t) with inference for categorical proportions (which uses z). Correct move: Only use t for inference about quantitative data (means); always use z for inference for categorical data (proportions).
• Wrong move: Claiming the t distribution is only appropriate when $n\ge 30$. Why: Students confuse the Central Limit Theorem normality condition with the requirement to use t. Correct move: Use t whenever $\sigma$ is unknown, regardless of sample size, as long as the normality condition is met (for small n, check no strong skewness/outliers).

6. Practice Questions (AP Statistics Style)

Question 1 (Multiple Choice)

Which of the following correctly describes a difference between the t distribution and the standard normal (z) distribution? A) The t distribution is skewed right, while the z distribution is symmetric. B) The t distribution has lighter tails than the z distribution for any finite degrees of freedom. C) As degrees of freedom increase, the t distribution approaches the z distribution. D) The t distribution is centered at a value that depends on degrees of freedom, while the z distribution is always centered at 0.

Worked Solution: We can eliminate incorrect options one by one. Option A is wrong: both the t and z distributions are symmetric. Option B is wrong: the t distribution has heavier, not lighter, tails than z for finite degrees of freedom, to account for extra uncertainty from estimating $\sigma$. Option D is wrong: the t distribution is always centered at 0, just like the z distribution. Only Option C is correct: as sample size (and thus degrees of freedom) increases, $s$ becomes a more accurate estimate of $\sigma$, so the t distribution converges to the z distribution. Correct answer: C.

Question 2 (Free Response)

A student wants to estimate the mean number of hours full-time college students spend studying per week at their campus. They collect a random sample of 15 full-time students, and find a sample mean of 18.2 hours and a sample standard deviation of 4.8 hours. The population standard deviation is unknown. (a) What degrees of freedom should be used for a t procedure to estimate the population mean? Justify your calculation. (b) A classmate claims that because the sample size is less than 30, the t distribution cannot be used here. Do you agree? Explain why or why not. (c) How would the critical value $t^{\ast }$ for 95% confidence change if the student added 10 more students to their random sample, keeping the confidence level the same? Justify your answer.

Worked Solution: (a) For a one-sample t procedure for a population mean, degrees of freedom follow the formula $df=n-1$, where $n$ is the sample size. Here, $n=15$, so $df=15-1=14$. We subtract 1 because we lose one degree of freedom when we use the sample mean to estimate the population mean for calculating the sample standard deviation. (b) I disagree. The $n\ge 30$ rule is for the normality condition via the Central Limit Theorem, not a requirement for using the t distribution. For small samples, we can use t procedures if the population distribution is approximately normal. The sample is random, and no extreme outliers or strong skewness are mentioned in the problem, so conditions for using t are satisfied. (c) Adding 10 students increases the sample size from 15 to 25, so degrees of freedom increases from 14 to 24. As degrees of freedom increase, the t distribution becomes less spread out, so the 95% critical value $t^{\ast }$ will decrease. A larger sample reduces uncertainty, so a smaller critical value is needed for the same confidence level.

Question 3 (Application / Real-World Style)

A nutrition researcher wants to test whether the mean calorie content of craft beer pints at local breweries is greater than the 180 calorie standard for mass-produced beer. They collect a random sample of 27 craft beer pints from local breweries, and do not know the population standard deviation of calorie content. (a) Is the t distribution appropriate for this study? If so, what degrees of freedom should be used? (b) The researcher calculates a t test statistic of 1.78. The one-tailed critical value for $\alpha =0.05$ with the correct df is 1.706. What conclusion can the researcher draw in context?

Worked Solution: (a) Check conditions for using the t distribution: the sample is random (given), $\sigma$ is unknown (given), and the sample size is $n=27$. Even though $n<30$, no extreme skewness or outliers are mentioned, so the normality condition is satisfied. The t distribution is appropriate. Degrees of freedom are $df=n-1=27-1=26$. (b) The calculated test statistic of 1.78 is greater than the critical value of 1.706. This means there is sufficient evidence at the $\alpha =0.05$ significance level to conclude that the mean calorie content of craft beer pints at local breweries is greater than the 180 calorie standard for mass-produced beer.

7. Quick Reference Cheatsheet

8. What's Next

Understanding the t distribution is the foundational prerequisite for all inference procedures for population means, which are the core of Unit 7. Next you will apply the properties of the t distribution to construct confidence intervals for one and two population means, and conduct hypothesis tests for one and two population means. Without mastering how to calculate degrees of freedom, identify when to use t instead of z, and interpret t distribution shape, you will not be able to correctly calculate critical values, p-values, or intervals, leading to lost points on both MCQ and FRQ. Beyond Unit 7, the t distribution is also used for inference for the slope of a regression line in Unit 9, so this concept appears across multiple sections of the AP exam.

• Confidence Intervals for a Population Mean
• Hypothesis Tests for a Population Mean
• Inference for the Slope of a Regression Model
• Comparing Two Population Means