Constructing a Confidence Interval for a Population Mean - AP Statistics

SECTION 1

Historical Context & Motivation

The fundamental challenge of statistics has always been the same: how do we use limited information from a sample to say something reliable about a much larger population? Before the twentieth century, scientists and researchers often reported single-number estimates—point estimates—without any formal way to communicate how uncertain those estimates were. The development of confidence intervals gave statisticians a principled framework for quantifying the precision of an estimate, fundamentally transforming how empirical research communicates its findings.

1908 • Student's t-Distribution

William Sealy Gosset, publishing under the pseudonym "Student" while working at the Guinness Brewery, derived the t-distribution for small samples. His work addressed the practical reality that the population standard deviation is almost never known.

1934 • Neyman's Confidence Interval Framework

Jerzy Neyman formalized the concept of a confidence interval, establishing the repeated-sampling interpretation: in the long run, a stated percentage of such intervals will contain the true parameter. This framework remains the standard in frequentist statistics.

1937 • Fisher–Neyman Debate

R. A. Fisher and Neyman clashed over the interpretation of interval estimation. Fisher preferred "fiducial" intervals, while Neyman's confidence interval approach ultimately became the dominant paradigm taught worldwide.

2000s • Modern Emphasis on Interval Estimation

Major journals such as the APA Publication Manual began requiring confidence intervals alongside hypothesis tests, reflecting a broader movement toward reporting effect sizes and estimation rather than relying solely on p-values.

The central question that confidence intervals address is deceptively simple: given a sample mean x̄, what range of values for the true population mean μ is plausible? Answering this question requires understanding sampling variability, the shape of the sampling distribution, and how to account for the fact that we must estimate the population standard deviation from the sample itself. This section of the lesson develops each of those ideas and shows you how to construct, interpret, and apply confidence intervals for a population mean—skills that are essential for the AP Statistics exam.

SECTION 2

Core Principles & Definitions

Before constructing a confidence interval, you need a clear understanding of the underlying concepts. A confidence interval is a range of plausible values for a population parameter, calculated from sample data, together with a stated confidence level that describes the long-run capture rate of the procedure. The interval takes the general form: point estimate ± margin of error. Each component of this formula carries specific statistical meaning that you must be able to articulate on the AP exam.

Point Estimate (x̄)

The sample mean x̄ serves as the best single-value estimate of the unknown population mean μ. It is the center of the confidence interval and comes directly from the data you collect.

Margin of Error

The margin of error equals the critical value multiplied by the standard error: t* × (s / √n). It quantifies how far the interval extends above and below x̄ and captures the uncertainty due to sampling variability.

Confidence Level (C)

The confidence level (commonly 90%, 95%, or 99%) describes the proportion of intervals that would capture μ if the sampling procedure were repeated many times under identical conditions. It does not describe the probability that one particular interval contains μ.

Standard Error (SE)

The standard error of the sample mean is s / √n, where s is the sample standard deviation and n is the sample size. It estimates the standard deviation of the sampling distribution of x̄ and shrinks as n increases.

Critical Value (t*)

The critical value t* comes from the t-distribution with n − 1 degrees of freedom and corresponds to the desired confidence level. For a 95% confidence interval with large n, t* ≈ 1.96, approaching the z* value.

✦

KEY TAKEAWAY

Think of a confidence interval like casting a net to catch a fish whose exact location you cannot see. The sample mean tells you roughly where to throw the net (point estimate), and the margin of error determines how wide you make the net. A higher confidence level means a wider net—more likely to catch the fish, but less precise about exactly where the fish is. The key insight is that the net's width depends on both how variable the water is (standard error) and how certain you want to be (confidence level).

SECTION 3

Visualizing the Confidence Interval

A confidence interval is best understood visually by examining the sampling distribution of x̄. If we could draw every possible sample of size n from the population, the distribution of those sample means would be approximately normal (by the Central Limit Theorem) with center μ and standard deviation σ/√n. Because we rarely know σ, we substitute s and use the t-distribution, which is slightly wider in the tails to account for this extra uncertainty. The diagram below illustrates how the middle C% of the t-distribution defines the critical values that determine the margin of error.

The bell curve represents the sampling distribution of x̄ centered at the true population mean μ. The shaded cyan region captures the middle 95% of sample means, bounded by the critical values −t* and +t*. The pink tails each contain 2.5% of the distribution. The amber segment at the bottom shows the resulting confidence interval: x̄ ± t*·(s/√n).

Notice that as the confidence level increases from 95% to 99%, the critical value t* grows, pushing the dashed pink boundary lines outward and making the interval wider. Conversely, increasing the sample size n reduces the standard error s/√n, which narrows the interval without sacrificing confidence. This trade-off between confidence and precision is central to the design of statistical studies and is a frequent topic on the AP exam.

SECTION 4

Mathematical Framework

The construction of a confidence interval for a population mean relies on the one-sample t-interval procedure. This procedure is used when the population standard deviation σ is unknown—which is nearly always the case in practice—and the sampling distribution of x̄ can be reasonably approximated by a t-distribution. The formulas below constitute the core machinery you need.

CONFIDENCE INTERVAL FOR μ

x̄ ± t* · (s / √n)

x̄ = sample mean, t* = critical value from the t-distribution with df = n − 1, s = sample standard deviation, n = sample size. The quantity t* · (s / √n) is the margin of error.

STANDARD ERROR OF x̄

SE = s / √n

The standard error estimates the standard deviation of the sampling distribution of x̄. As n increases, SE decreases proportionally to the square root of n, reflecting increased precision from larger samples.

DEGREES OF FREEDOM

df = n − 1

The degrees of freedom control the shape of the t-distribution. Smaller df values yield heavier tails (wider intervals), compensating for the added uncertainty when s estimates σ from fewer observations.

Conditions for Inference

Before applying the t-interval formula, three conditions must be verified. On the AP exam, checking these conditions explicitly is essential for earning full credit on free-response questions.

✓ Random: The data must come from a random sample or a randomized experiment. This ensures the sample is representative and allows the results to generalize to the population.
✓ Independence (10% condition): Individual observations must be independent. When sampling without replacement, the sample size n should be no more than 10% of the population size N, i.e., n ≤ 0.10N.
✓ Normal/Large Sample: The sampling distribution of x̄ must be approximately normal. This is satisfied if the population distribution is roughly normal, or if n ≥ 30 (by the Central Limit Theorem). For smaller samples, check a dotplot, histogram, or normal probability plot for strong skewness or outliers.

SECTION 5

Choosing Between z and t, and the Effect of Sample Size

Students often ask when to use a z-interval versus a t-interval. In practice, the answer for the AP Statistics exam is almost always the t-interval, because the population standard deviation σ is rarely known. The z-interval (using z* and σ) is reserved for the unusual case where σ is given in the problem statement. The following table clarifies the decision and highlights the key differences between the two procedures.

Feature	z-Interval	t-Interval
Population σ	Known	Unknown; estimated by s
Critical value	z* from standard normal (e.g., 1.96 for 95%)	t* from t-distribution with df = n − 1
Interval formula	x̄ ± z* · (σ / √n)	x̄ ± t* · (s / √n)
Tail behavior	Lighter tails	Heavier tails (wider interval for small n)
AP exam usage	Rare; only when σ is explicitly provided	Default procedure for inference about μ

Comparison of z-interval and t-interval procedures for a population mean

As the sample size increases, the margin of error decreases at a rate proportional to 1/√n. Quadrupling the sample size from 10 to 40 cuts the margin of error roughly in half (from 9.3 to 4.6). This diminishing-returns pattern means that extremely large samples yield only modest additional precision.

💡

AP Exam Tip

When a free-response question asks you to construct a confidence interval, you must explicitly state all three conditions (Random, Independence/10%, Normal/Large Sample), name the procedure (one-sample t-interval), show the formula with substitution, give the interval, and interpret it in context. Omitting any of these steps typically costs points.

SECTION 6

Worked Example

A nutritionist wants to estimate the mean daily caloric intake of college students at a large university. She takes a simple random sample of 36 students and records their caloric intake over one day. The sample yields x̄ = 2,180 calories and s = 240 calories. Construct a 95% confidence interval for the population mean daily caloric intake μ.

One-Sample t-Interval for μ

1
Step 1 — Check Conditions

Random: The problem states that a simple random sample was used. ✓ Independence (10% condition): A large university has well over 360 students, so n = 36 ≤ 10% of the population. ✓ Normal/Large Sample: n = 36 ≥ 30, so by the Central Limit Theorem the sampling distribution of x̄ is approximately normal. ✓
2
Step 2 — Identify the Procedure and Parameters

Because σ is unknown and we are estimating a single population mean, we use the one-sample t-interval. We have x̄ = 2180, s = 240, n = 36, and df = 36 − 1 = 35.
3
Step 3 — Find the Critical Value t*

For a 95% confidence level and df = 35, the t-table (or calculator invT function) gives t* = 2.030.

t* = 2.030
4
Step 4 — Calculate the Standard Error and Margin of Error

SE = s / √n = 240 / √36 = 240 / 6 = 40. Margin of Error = t* × SE = 2.030 × 40 = 81.2.

ME = 81.2 calories
5
Step 5 — Construct the Interval

x̄ ± ME = 2180 ± 81.2, which gives (2098.8, 2261.2).

95% CI: (2098.8, 2261.2) calories
6
Step 6 — Interpret in Context

We are 95% confident that the true mean daily caloric intake of all college students at this university is between 2,098.8 and 2,261.2 calories. This means that if we were to repeat this sampling procedure many times, approximately 95% of the resulting intervals would capture the true population mean μ.

⚠

Interpretation Warning

Do NOT say "there is a 95% probability that μ is in this interval." Once the interval is calculated, μ is either in it or it is not—there is no probability involved for a specific interval. The 95% refers to the long-run success rate of the method, not the probability for any single interval. This distinction is a common source of lost points on the AP exam.

SECTION 7

Strengths, Limitations & Common Errors

Strengths	Limitations
Provides a range of plausible values rather than a single point estimate, communicating the precision of the estimate	The stated confidence level is only valid when the conditions (random, independence, normality) are met; violations can distort the capture rate
The t-distribution correctly accounts for extra uncertainty when σ is unknown and estimated by s	Sensitive to outliers: extreme values inflate s, which widens the interval and may distort the results for small samples
Easy to compute with a calculator (TInterval function on TI-84) and straightforward to interpret in context	Wider interval does not always mean better; there is a trade-off between confidence and precision that requires careful study design
Robust for large samples: by the CLT, the t-interval performs well even if the population distribution is somewhat skewed when n ≥ 30	Cannot compensate for bias in data collection; a confidence interval from a convenience sample has no guaranteed interpretation

Strengths and limitations of the one-sample t-interval

✦

KEY TAKEAWAY

A confidence interval is only as trustworthy as the data-collection process behind it. Much like a thermometer that is perfectly precise but poorly calibrated will give you consistently wrong readings, a confidence interval built on biased data will consistently miss the true parameter—no matter how narrow or how high the confidence level. Always verify the conditions before interpreting the interval, and remember that the interval quantifies sampling variability, not systematic error.

SECTION 8

Connection to Hypothesis Testing and Advanced Methods

Confidence intervals and hypothesis tests are two sides of the same inferential coin. A 95% confidence interval for μ contains all values of μ₀ that would not be rejected by a two-sided significance test at α = 0.05. If a hypothesized value μ₀ falls inside the confidence interval, you would fail to reject H₀: μ = μ₀ at that significance level; if μ₀ falls outside, you would reject. This duality means that confidence intervals convey strictly more information than hypothesis tests—they show not just whether a value is plausible, but the entire range of plausible values.

Feature	One-Sample t-Interval (This Lesson)	Two-Sample t-Interval (Advanced)
Parameter estimated	μ (one population mean)	μ₁ − μ₂ (difference between two population means)
Standard error	s / √n	√(s₁²/n₁ + s₂²/n₂)
Degrees of freedom	n − 1	Welch's approximation (calculator computes)
Number of samples	1	2 independent samples
Conditions	Random, 10% rule, Normal/Large Sample	Same conditions checked for each sample independently

Comparison of one-sample and two-sample t-intervals

Beyond the AP curriculum, confidence intervals extend to paired t-intervals (for matched-pairs designs), regression slope intervals, and even bootstrap confidence intervals that do not assume normality at all. Mastering the one-sample t-interval gives you a template that generalizes: every confidence interval follows the pattern point estimate ± (critical value) × (standard error). The specific formulas for the standard error and the distribution used for the critical value change, but the logic and structure remain the same.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

A researcher constructs a 95% confidence interval for the mean weight of a species of bird and obtains (38.2 g, 42.8 g). Which of the following is a correct interpretation of this interval?

A. There is a 95% probability that the true mean weight of this bird species is between 38.2 g and 42.8 g.
B. We are 95% confident that the true mean weight of this bird species is between 38.2 g and 42.8 g.
C. 95% of all individual birds of this species weigh between 38.2 g and 42.8 g.
D. If we took one more sample, there is a 95% chance that the sample mean would fall between 38.2 g and 42.8 g.
E. 95% of all sample means from samples of the same size will be between 38.2 g and 42.8 g.

Show answer (B)

B is correct. The proper interpretation uses the phrase "we are [C]% confident that the true [parameter] is between [lower] and [upper]." This language reflects the long-run capture rate of the procedure. A is wrong because once the interval is computed, μ is either in it or not—there is no probability statement about a fixed parameter. C is wrong because the interval estimates the population mean, not the spread of individual values. D is wrong because the interval is about the parameter μ, not about future sample means. E is wrong because, while related to sampling distributions, the specific interval (38.2, 42.8) is not the region that captures 95% of sample means; it is the region that we believe captures μ.

Question 2: BASIC CALCULATION

A random sample of 25 light bulbs has a mean lifetime of x̄ = 1,200 hours with s = 100 hours. Using t* = 2.064 for a 95% confidence interval with 24 degrees of freedom, what is the margin of error?

A. 8.26 hours
B. 20.64 hours
C. 39.20 hours
D. 41.28 hours
E. 206.40 hours

Show answer (D)

D is correct. ME = t* × (s / √n) = 2.064 × (100 / √25) = 2.064 × (100 / 5) = 2.064 × 20 = 41.28 hours. A results from dividing t* by n instead of dividing s by √n. B results from dividing t* × s by n instead of √n. C uses z* = 1.96 instead of t* = 2.064. E multiplies t* by s without dividing by √n at all.

Question 3: INTERMEDIATE

A 90% confidence interval for a population mean is (52.3, 61.7). If the researcher wants to construct a 99% confidence interval from the same sample data, how will the new interval compare to the original?

A. The new interval will be narrower and centered at the same value.
B. The new interval will be wider and centered at the same value.
C. The new interval will be wider and centered at a different value.
D. The new interval will be the same width but centered at a different value.
E. The effect on the interval cannot be determined without knowing the sample size.

Show answer (B)

B is correct. Increasing the confidence level from 90% to 99% increases the critical value t*, which increases the margin of error. Since x̄, s, and n remain unchanged, the center of the interval (x̄ = (52.3 + 61.7)/2 = 57.0) stays the same—only the width changes. A is wrong because higher confidence requires a wider, not narrower, interval. C is wrong because the center is always x̄ regardless of the confidence level. D is wrong because the width must change when the confidence level changes. E is wrong because the direction of the effect (wider) is guaranteed regardless of n; only the magnitude depends on n.

Question 4: APPLIED

4 pts

A city transportation department wants to estimate the mean commute time (in minutes) for workers who use public transit. A random sample of 50 commuters yields x̄ = 34.6 minutes and s = 11.2 minutes. (a) Verify that the conditions for constructing a one-sample t-interval are satisfied. (b) Construct a 95% confidence interval for the mean commute time μ. (c) The department claims the mean commute time is 30 minutes. Based on your interval, is there convincing evidence that the true mean commute time differs from 30 minutes? Explain. (d) If the department wanted to reduce the margin of error to 2 minutes while maintaining 95% confidence, approximately how large a sample would be needed? (Use z* = 1.96 as an approximation for planning purposes.)

Show answer

Scoring Rubric (4 points): Part (a) — 1 point: Random: stated that the sample is random. Independence: 50 commuters is certainly less than 10% of all public transit commuters in a city. Normal/Large Sample: n = 50 ≥ 30, so the CLT guarantees the sampling distribution of x̄ is approximately normal. Award 1 point for correctly checking all three conditions with context. Part (b) — 1 point: Identify one-sample t-interval with df = 49. t* ≈ 2.010. SE = 11.2/√50 ≈ 1.584. ME = 2.010 × 1.584 ≈ 3.18. Interval: 34.6 ± 3.18 = (31.42, 37.78) minutes. Award 1 point for correct interval with work shown. Part (c) — 1 point: Because 30 minutes is not contained in the 95% confidence interval (31.42, 37.78), there is convincing evidence that the true mean commute time differs from 30 minutes. At the 5% significance level, we would reject H₀: μ = 30. Award 1 point for correct conclusion linked to whether 30 is in the interval, stated in context. Part (d) — 1 point: Using the planning formula n = (z* × s / ME)² = (1.96 × 11.2 / 2)² = (10.976)² ≈ 120.5. Rounding up, the department would need at least n = 121 commuters. Award 1 point for correct setup and answer, rounded up to the next whole number.

Question 5: CRITICAL THINKING

4 pts

A medical researcher collected a random sample of 12 patients' recovery times (in days) after a particular surgery. The data are: 5, 7, 8, 8, 9, 9, 10, 10, 11, 12, 14, 42. The sample mean is x̄ = 12.08 and the sample standard deviation is s = 9.82. (a) Explain why the conditions for a one-sample t-interval may not be fully satisfied for these data. Identify the specific concern. (b) A 95% t-interval computed from these statistics is (5.84, 18.33). Explain why this interval may not be trustworthy, even though the formula was applied correctly. (c) Suggest a statistical strategy the researcher could use to produce a more reliable confidence interval, and explain why your strategy would help. (d) Suppose the researcher removes the outlier (42 days) and recalculates: x̄ = 9.36, s = 2.42, n = 11. Construct a 95% confidence interval using t* = 2.228 and compare it to the original interval. What does this comparison reveal about the influence of outliers?

Show answer

Scoring Rubric (4 points): Part (a) — 1 point: The Normal/Large Sample condition is not satisfied. With n = 12 (less than 30), we need the population distribution to be approximately normal. The value of 42 is a clear outlier (more than 3 standard deviations above the mean of the other 11 values), indicating strong right skewness. The t-procedures are not robust against outliers with small sample sizes. Award 1 point for identifying the outlier and connecting it to the normality condition with n < 30. Part (b) — 1 point: The outlier inflates both x̄ (pulling the center upward) and s (making the spread appear much larger than it would be without the outlier). Because the t-interval assumes the sampling distribution of x̄ is approximately t-distributed, the actual capture rate of the interval may differ substantially from the nominal 95%. The interval is very wide (width ≈ 12.5 days) largely because of one extreme observation, so it may not accurately reflect the typical variability in recovery times. Award 1 point for explaining that the outlier distorts both the center and spread, undermining the validity of the stated confidence level. Part (c) — 1 point: Acceptable strategies include: (1) collecting a larger sample so the CLT applies, (2) using a nonparametric method such as a bootstrap confidence interval that does not assume normality, or (3) applying a transformation (e.g., log transformation) to reduce skewness before constructing the interval. Any one strategy with a correct explanation of why it mitigates the outlier/normality concern earns the point. Part (d) — 1 point: SE = 2.42/√11 ≈ 0.730. ME = 2.228 × 0.730 ≈ 1.626. Interval: 9.36 ± 1.63 = (7.73, 10.99) days. This interval is dramatically narrower (width ≈ 3.26 days versus ≈ 12.49 days) and centered much lower (9.36 vs. 12.08). This demonstrates that a single outlier can inflate the sample mean, inflate s, and produce a misleadingly wide confidence interval. However, simply removing an outlier must be justified on subject-matter grounds (e.g., the 42-day case may represent a complication) rather than purely for statistical convenience. Award 1 point for the correct interval, a comparison with the original, and acknowledgment of the outlier's influence.

SUMMARY

Lesson Summary

Summary

A confidence interval for a population mean takes the form x̄ ± t* · (s / √n), where x̄ is the sample mean, t* is the critical value from the t-distribution with df = n − 1, and s / √n is the standard error. Before applying this formula, you must verify three conditions: Random sampling, Independence (10% condition), and Normal/Large Sample. On the AP exam, you must explicitly check these conditions, name the procedure, show your calculations, and interpret the interval in context.

The confidence level describes the long-run capture rate of the procedure—not the probability that a specific interval contains μ. Increasing the confidence level widens the interval; increasing the sample size narrows it by reducing the standard error. The one-sample t-interval is the default procedure when σ is unknown, and it serves as the foundational template for all confidence interval procedures you will encounter in AP Statistics: point estimate ± (critical value) × (standard error).