What you'll learn
This guide covers statistical inference for quantitative data when working with means. You'll learn how to construct and interpret confidence intervals for a population mean, conduct one-sample and two-sample t-tests, and verify the conditions required for valid inference. These methods are essential for making conclusions about populations based on sample data.
Key terms and definitions
Standard error — the estimated standard deviation of a sampling distribution, calculated from sample data; for a sample mean, SE = s/√n where s is the sample standard deviation
t-distribution — a family of symmetric, bell-shaped distributions with heavier tails than the normal distribution, used when the population standard deviation is unknown; degrees of freedom determine the specific distribution
Confidence interval — a range of plausible values for a population parameter, constructed from sample data, with an associated confidence level indicating the long-run success rate of the method
Margin of error — half the width of a confidence interval, calculated as the critical value multiplied by the standard error
Null hypothesis (H₀) — a statement of no effect or no difference, typically claiming a population parameter equals a specific value
Alternative hypothesis (Hₐ) — the research claim we seek evidence for, stating the population parameter is different from, greater than, or less than the null value
p-value — the probability of obtaining sample results at least as extreme as those observed, assuming the null hypothesis is true
Degrees of freedom — the number of independent pieces of information available to estimate variability; for a one-sample t-procedure, df = n - 1
Core concepts
Conditions for inference about means
Before conducting any inference procedure for means, you must verify three conditions:
Random condition: The data must come from a random sample or randomized experiment. This ensures the sample is representative of the population and allows generalization.
10% condition: When sampling without replacement, the sample size should be no more than 10% of the population size (n ≤ 0.10N). This ensures independence of observations.
Normal/Large Sample condition: Either the population distribution is normal, or the sample size is large enough (generally n ≥ 30) for the Central Limit Theorem to apply. For smaller samples, check the data for strong skewness or outliers using a graph (histogram, boxplot, or normal probability plot).
When any condition is seriously violated, t-procedures may not be valid. Always state and check all three conditions before proceeding with inference.
One-sample t-confidence interval for a mean
A confidence interval for a population mean μ when σ is unknown uses the formula:
x̄ ± t × (s/√n)*
where:
- x̄ is the sample mean
- t* is the critical value from the t-distribution with df = n - 1
- s is the sample standard deviation
- n is the sample size
The confidence level (e.g., 95%, 90%, 99%) represents the percentage of all possible samples that would produce intervals capturing the true parameter. A 95% confidence level means that if we repeated the sampling process many times, approximately 95% of the intervals constructed would contain the true population mean.
To interpret a confidence interval correctly, state: "We are [C]% confident that the true mean [context] is between [lower bound] and [upper bound]." Never say there's a C% probability that μ is in the interval—the parameter is fixed, not random.
One-sample t-test for a mean
A one-sample t-test evaluates whether sample data provide convincing evidence that a population mean differs from a hypothesized value.
Test statistic formula: t = (x̄ - μ₀)/(s/√n)
where μ₀ is the hypothesized population mean.
The test statistic measures how many standard errors the sample mean is from the hypothesized value. Compare the t-statistic to the t-distribution with df = n - 1 to find the p-value.
Steps for conducting a hypothesis test:
State hypotheses: Define H₀ and Hₐ in context
- H₀: μ = μ₀
- Hₐ: μ ≠ μ₀ (two-sided), μ > μ₀ (upper-tailed), or μ < μ₀ (lower-tailed)
Identify the procedure and check conditions: Name the test (one-sample t-test) and verify Random, 10%, and Normal/Large Sample conditions
Calculate the test statistic and p-value: Use the formula above and technology or tables
Make a decision: Compare the p-value to the significance level α (commonly 0.05)
- If p-value < α: reject H₀
- If p-value ≥ α: fail to reject H₀
State a conclusion in context: Link the statistical decision to the research question
Two-sample t-procedures for comparing means
When comparing means from two independent groups, use two-sample t-procedures. These test whether μ₁ - μ₂ equals a specific value (usually 0) or estimate the difference with a confidence interval.
Additional conditions for two-sample procedures:
- Both samples must be independent (random samples or random assignment)
- The two groups must be independent of each other
- 10% condition for each sample
- Normal/Large Sample condition for each group
Two-sample t-confidence interval:
(x̄₁ - x̄₂) ± t* × √(s₁²/n₁ + s₂²/n₂)
Two-sample t-test statistic:
t = (x̄₁ - x̄₂) - (μ₁ - μ₂)₀ / √(s₁²/n₁ + s₂²/n₂)
The degrees of freedom for two-sample procedures use a complex formula (calculated by technology). A conservative approach uses df = min(n₁ - 1, n₂ - 1).
Paired t-procedures
When data come in natural pairs (before/after measurements, matched pairs, repeated measures on the same subjects), use paired t-procedures. Calculate the difference for each pair, then conduct a one-sample t-procedure on these differences.
Key distinction: Paired data have a natural pairing between observations in the two groups. Two-sample procedures require independent groups.
For paired data:
- Calculate differences: d = observation₁ - observation₂
- Check conditions on the differences
- Use one-sample t-procedures with d̄ (mean difference) and sₐ (standard deviation of differences)
The hypotheses for paired tests concern μₐ (the mean difference), not individual group means.
Interpreting computer output and using technology
AP Statistics exams often provide computer output rather than asking for hand calculations. You must be able to:
- Identify the correct procedure from output labels
- Extract the test statistic, degrees of freedom, and p-value
- Interpret confidence intervals from output
- Recognize when output matches one-sample, two-sample, or paired procedures
Standard output typically shows:
- Sample statistics (means, standard deviations, sample sizes)
- Standard error
- Test statistic (t-value)
- Degrees of freedom
- p-value (sometimes for two-sided test only)
- Confidence interval bounds
Worked examples
Example 1: One-sample t-confidence interval
Question: A marine biologist collected a random sample of 25 Caribbean spiny lobsters and measured their carapace length (in mm). The sample mean was 78.4 mm with a standard deviation of 6.2 mm. Construct and interpret a 95% confidence interval for the mean carapace length of all Caribbean spiny lobsters in this population.
Solution:
Identify procedure: One-sample t-interval for μ
Check conditions:
- Random: Stated as a random sample ✓
- 10%: 25 lobsters is less than 10% of all Caribbean spiny lobsters ✓
- Normal/Large Sample: n = 25 < 30, so we need to assume the distribution of carapace lengths is approximately normal or check data for strong skewness (assumed acceptable if not shown otherwise) ✓
Calculate:
- x̄ = 78.4 mm, s = 6.2 mm, n = 25
- df = 25 - 1 = 24
- t* = 2.064 (from t-table for 95% confidence, df = 24)
- SE = 6.2/√25 = 1.24
- Margin of error = 2.064 × 1.24 = 2.56
- Interval: 78.4 ± 2.56 = (75.84, 80.96) mm
Interpretation: We are 95% confident that the true mean carapace length of Caribbean spiny lobsters in this population is between 75.84 mm and 80.96 mm.
Example 2: Two-sample t-test
Question: A study compared the effectiveness of two teaching methods for mathematics in UK secondary schools. A random sample of 30 students used Method A (traditional instruction) and 28 students used Method B (technology-enhanced instruction). After 6 weeks, students took the same assessment. Method A had a mean score of 72.3 (s = 8.4), and Method B had a mean score of 76.8 (s = 7.9). Test whether there is convincing evidence that Method B produces higher mean scores than Method A at the α = 0.05 level.
Solution:
State hypotheses:
- H₀: μ_A - μ_B = 0 (no difference in mean scores)
- Hₐ: μ_A - μ_B < 0 (Method B has higher mean score)
Identify procedure and check conditions: Two-sample t-test for μ_A - μ_B
- Random: Random samples stated ✓
- Independent groups: Different students in each group ✓
- 10%: Both samples < 10% of all UK secondary students ✓
- Normal/Large Sample: Both n_A = 30 and n_B = 28 are close to 30, satisfying the condition ✓
Calculate:
- Test statistic: t = (72.3 - 76.8) / √(8.4²/30 + 7.9²/28)
- t = -4.5 / √(2.352 + 2.229) = -4.5 / 2.14 = -2.10
- df ≈ 55.7 (from technology) or conservative df = 27
- p-value ≈ 0.020 (one-tailed, from technology)
Decision: Since p-value (0.020) < α (0.05), reject H₀
Conclusion: There is convincing evidence that Method B (technology-enhanced instruction) produces a higher mean assessment score than Method A (traditional instruction).
Example 3: Paired t-test
Question: Ten students measured their resting pulse rate, then exercised vigorously for 5 minutes and measured their pulse rate immediately after. The differences (after - before) had a mean of 42.3 beats per minute with a standard deviation of 8.6 bpm. Test whether exercise significantly increases pulse rate (α = 0.01).
Solution:
State hypotheses:
- H₀: μ_d = 0 (no mean change in pulse rate)
- Hₐ: μ_d > 0 (mean pulse rate increases after exercise)
Identify procedure and check conditions: Paired t-test for μ_d
- Random: Not stated, but assume representative of typical response ⚠
- 10%: 10 students < 10% of all students ✓
- Normal/Large Sample: n = 10 is small; would need to check normality of differences with a plot ⚠
Calculate:
- d̄ = 42.3, s_d = 8.6, n = 10
- t = (42.3 - 0)/(8.6/√10) = 42.3/2.72 = 15.55
- df = 9
- p-value < 0.0001 (extremely small for such a large t-statistic)
Decision: Since p-value < 0.01, reject H₀
Conclusion: There is very strong evidence that exercise increases mean pulse rate in students.
Common mistakes and how to avoid them
• Confusing paired and two-sample procedures: Always identify whether data are paired (natural matching between observations) or come from two independent groups. Drawing a diagram of the data structure helps.
• Wrong interpretation of confidence intervals: Never say "there is a 95% probability that μ is in the interval." The correct interpretation refers to the method's long-run success rate: "We are 95% confident..."
• Stating conclusions without context: Always relate conclusions to the real-world situation. Don't just say "reject H₀"—explain what this means for the population being studied.
• Failing to check conditions properly: State all three conditions explicitly and verify each one. Saying "conditions are met" without justification earns no credit. Reference the problem context when checking.
• Confusing p-value with significance level: The p-value is calculated from data; the significance level α is chosen before conducting the test. Compare p-value to α to make decisions.
• Using z-procedures when σ is unknown: When the population standard deviation is unknown (almost always), use t-procedures, not z-procedures. The t-distribution accounts for additional uncertainty from estimating σ with s.
Exam technique for "Inference for Quantitative Data: Means"
• For "test the claim" questions: Follow all five steps of hypothesis testing. State hypotheses with proper notation and context, name the procedure, verify conditions with justification, calculate the test statistic and p-value, make a decision, and state a conclusion in context. Full solutions typically earn 4-5 marks.
• For "construct and interpret" confidence interval questions: Always check conditions first (1-2 marks), show the calculation or identify correct output (1-2 marks), and provide a proper interpretation in context (1 mark). The interpretation must include the confidence level, parameter being estimated, and context.
• Command word "justify": When asked to justify a choice of procedure (paired vs. two-sample), explain the data structure specifically. Generic statements earn no credit.
• Show working efficiently: On calculator-allowed exams, you may use technology for calculations, but must show the formula setup and values being used. Write: "t = (x̄ - μ₀)/(s/√n) = (78.4 - 75)/(6.2/√25) = 2.74" not just "t = 2.74".
Quick revision summary
Inference for means uses t-procedures when σ is unknown. Always verify three conditions: Random sampling, 10% rule, and Normal/Large Sample. One-sample procedures test claims about or estimate a single population mean. Two-sample procedures compare means from independent groups. Paired procedures analyze differences within matched pairs. Confidence intervals estimate parameters; hypothesis tests evaluate claims. Interpret results in context, linking statistical decisions to real-world conclusions. The p-value determines whether evidence against H₀ is convincing at significance level α.