CIs & Hypothesis Testing — AHL 4.9 Confidence intervals and...

SECTION 1

Historical Context & Motivation

Imagine you want to know the average height of every 16-year-old in your country. Measuring every single person would be impossible, so instead you measure a sample and try to draw conclusions about the whole population. But how confident can you be in those conclusions? This question drove statisticians for over a century to develop the tools of confidence intervals and hypothesis testing — two of the most powerful ideas in modern statistics.

1908

Student's t-Distribution

William Sealy Gosset, working at the Guinness brewery under the pen name "Student," developed the t-distribution to handle small samples. His work made it possible to draw conclusions even when data sets were limited.

1925

Fisher's Significance Testing

Ronald Fisher formalized the concept of statistical significance and introduced the p-value as a measure of evidence against a null hypothesis. His 5% significance level became a widely adopted standard.

1933

Neyman–Pearson Framework

Jerzy Neyman and Egon Pearson introduced the formal hypothesis testing framework with null and alternative hypotheses, Type I and Type II errors, and the concept of confidence intervals as we use them today.

1990s–present

Technology Revolution

Graphing calculators and statistical software (like GDCs, GeoGebra, and SPSS) made it possible for students and researchers to perform complex hypothesis tests and construct confidence intervals rapidly, shifting the focus from hand computation to interpretation.

Today, the IB Mathematics: Applications and Interpretation course expects you to harness technology — your GDC (graphic display calculator) or statistical software — to construct confidence intervals and perform hypothesis tests efficiently. The central question this topic addresses is: Given limited sample data, how can we make reliable, quantified claims about an entire population?

SECTION 2

Core Principles & Definitions

Before diving into calculations, you need a solid grasp of the foundational vocabulary and ideas. Confidence intervals and hypothesis tests are two sides of the same coin — both use sample statistics to make inferences about population parameters. A parameter is a fixed but usually unknown number describing the whole population (such as the true population mean μ), while a statistic is a value calculated from your sample (such as the sample mean x̄).

Confidence Interval (CI)

A range of values, computed from sample data, that is likely to contain the true population parameter. A 95% CI means that if you repeated the sampling process many times, about 95% of the intervals you calculate would capture the true value.

Confidence Level

The percentage (commonly 90%, 95%, or 99%) expressing how often the interval procedure would capture the true parameter across repeated samples. A higher confidence level produces a wider interval.

Null Hypothesis (H₀)

A statement of 'no effect' or 'no difference' that we assume to be true until evidence suggests otherwise. For example, H₀: μ = 50 claims the population mean is 50.

Alternative Hypothesis (H₁)

The claim we are trying to find evidence for. It can be two-tailed (H₁: μ ≠ 50), left-tailed (H₁: μ < 50), or right-tailed (H₁: μ > 50), depending on the research question.

p-Value

The probability of obtaining sample results at least as extreme as the observed data, assuming H₀ is true. A small p-value (typically < 0.05) provides strong evidence against H₀.

✦ KEY TAKEAWAY

Think of a confidence interval like a fishing net. A wider net (higher confidence level) is more likely to catch the fish (the true parameter), but it's less precise about exactly where the fish is. Hypothesis testing is like a courtroom trial: we assume the defendant (H₀) is innocent until the evidence (p-value) is strong enough to convince us otherwise.

SECTION 3

Visualising Confidence Intervals

A confidence interval is built around the sample mean. The diagram below shows how a 95% confidence interval relates to the normal distribution. The interval stretches from the sample mean minus the margin of error to the sample mean plus the margin of error, capturing the central 95% of the sampling distribution.

The shaded cyan region represents the central 95% of the sampling distribution. The pink boundaries mark the lower and upper limits of the confidence interval. Each tail contains 2.5% of the area, giving a total of 5% outside the interval.

Notice that the confidence interval is centred on the sample mean x̄, not on the unknown population mean μ. The width of the interval depends on three things: the confidence level (which determines the z-value), the population standard deviation σ (or sample standard deviation s), and the sample size n. Larger samples produce narrower, more precise intervals because the denominator √n grows.

SECTION 4

Mathematical Framework

In the IB Applications and Interpretation HL course, you will use your GDC or technology to compute confidence intervals and p-values. However, understanding the formulas behind these calculations helps you interpret the results correctly and know when each test is appropriate.

Confidence Interval for a Population Mean

CONFIDENCE INTERVAL (σ KNOWN)

CI = x̄ ± z* × (σ / √n)

x̄ = sample mean, z* = critical z-value for the chosen confidence level (e.g., 1.96 for 95%), σ = population standard deviation, n = sample size. The term z* × (σ / √n) is called the margin of error.

CONFIDENCE INTERVAL (σ UNKNOWN — t-INTERVAL)

CI = x̄ ± t* × (s / √n)

When σ is unknown (which is almost always the case in practice), replace σ with the sample standard deviation s and use the t-distribution with (n − 1) degrees of freedom. Your GDC handles this automatically with a t-interval function.

Hypothesis Test for a Population Mean

TEST STATISTIC (z-TEST)

z = (x̄ − μ₀) / (σ / √n)

μ₀ = the hypothesised population mean (from H₀), x̄ = sample mean, σ = population standard deviation, n = sample size. This tells you how many standard errors the sample mean is away from the hypothesised value.

TEST STATISTIC (t-TEST)

t = (x̄ − μ₀) / (s / √n)

Used when σ is unknown. The t-statistic follows a t-distribution with df = n − 1. Your GDC computes both the test statistic and the p-value in one step.

💡 IB Exam Tip

On IB exams, you are expected to state H₀ and H₁ clearly, identify the test and significance level, report the p-value from your GDC, and write a conclusion in context. You do not need to compute test statistics by hand, but you do need to show your GDC setup (e.g., which inputs you entered).

SECTION 5

The Hypothesis Testing Decision Process

Hypothesis testing follows a structured, step-by-step process. The flowchart below shows the complete decision pathway from setting up the hypotheses to writing your final conclusion. Every IB question expects you to follow these steps in order.

The five-step hypothesis testing process. After stating your hypotheses and choosing a significance level α (commonly 0.05), you use your GDC to compute the p-value. If p < α, you reject H₀; otherwise, you do not reject H₀. Always write your conclusion in context of the original problem.

Common Significance Levels and z-Values

Relationship between confidence level, significance level, and critical z-values for two-tailed tests.

Confidence Level	Significance Level (α)	Critical z* (two-tailed)
90%	0.10	1.645
95%	0.05	1.960
99%	0.01	2.576

Notice the inverse relationship: as the confidence level increases, the significance level α decreases, and the critical z-value increases. This means you need stronger evidence (a more extreme sample) to reject H₀ at a higher confidence level. On the IB exam, if no significance level is stated, the default is α = 0.05 (5%).

SECTION 6

Worked Example: CI and Hypothesis Test Using Technology

A school nurse claims that the mean resting heart rate of students at the school is 72 beats per minute (bpm). A sample of 35 students is taken, yielding a sample mean of x̄ = 75.2 bpm and a sample standard deviation of s = 8.4 bpm. Test the nurse's claim at the 5% significance level using a two-tailed t-test, and construct a 95% confidence interval for the true mean heart rate.

Two-Tailed t-Test and 95% Confidence Interval

Step 1 — State the HypothesesThe nurse's claim is that the population mean heart rate is 72 bpm. Since we want to test whether the true mean is different from 72 (not specifically higher or lower), we set up a two-tailed test.

H₀: μ = 72 H₁: μ ≠ 72

Step 2 — Identify the Test and Significance LevelSince the population standard deviation σ is unknown (we only have the sample standard deviation s = 8.4), we use a t-test. The significance level is α = 0.05. The degrees of freedom are df = n − 1 = 35 − 1 = 34.

Two-tailed t-test, α = 0.05, df = 34

Step 3 — Enter Data into GDCOn your GDC (e.g., TI-84: STAT → TESTS → T-Test), enter: μ₀ = 72, x̄ = 75.2, s = 8.4, n = 35, and select the two-tailed alternative (μ ≠ μ₀). For the confidence interval, use STAT → TESTS → TInterval with the same x̄, s, n, and C-Level = 0.95.

GDC inputs: μ₀ = 72, x̄ = 75.2, s = 8.4, n = 35

Step 4 — Read the GDC OutputThe GDC returns: t = (75.2 − 72) / (8.4 / √35) = 3.2 / 1.4198… ≈ 2.254 and p-value ≈ 0.0307. For the confidence interval: 95% CI = (72.314, 78.086).

t ≈ 2.254, p-value ≈ 0.0307, 95% CI: (72.3, 78.1)

Step 5 — Make a Decision and Conclude in ContextSince the p-value (0.0307) is less than α (0.05), we reject H₀. Notice also that 72 is not inside the 95% confidence interval (72.3, 78.1), which is consistent with our rejection. Both methods agree.

There is sufficient evidence at the 5% significance level to reject the nurse's claim. The mean resting heart rate of students at this school is significantly different from 72 bpm.

🔗 CI and Hypothesis Test Agreement

For a two-tailed test at significance level α, if the hypothesised value μ₀ falls outside the (1 − α) × 100% confidence interval, you will reject H₀. If μ₀ falls inside the interval, you do not reject. This means a 95% CI and a two-tailed test at α = 0.05 always give the same conclusion.

SECTION 7

Strengths, Limitations & Common Errors

Confidence intervals and hypothesis tests are incredibly useful, but they are not foolproof. Understanding their strengths and limitations helps you avoid the most common mistakes IB students make on exams.

Key strengths and common pitfalls of confidence intervals and hypothesis testing.

Aspect	Strengths	Limitations / Common Errors
Interpretation of CI	Gives a plausible range for the parameter, not just a single estimate.	A 95% CI does NOT mean there is a 95% probability the true mean is in this specific interval. The true mean is either in it or not. The 95% refers to the long-run success rate of the method.
p-value	Quantifies the strength of evidence against H₀ on a continuous scale.	A p-value is NOT the probability that H₀ is true. It is the probability of getting data this extreme IF H₀ were true.
Conclusion language	Provides a structured, repeatable decision process.	Never say "accept H₀." You can only "fail to reject" or "do not reject" H₀. Absence of evidence is not evidence of absence.
Sample size	Larger n → narrower CI and more powerful tests.	With a very large sample, even a trivially small difference can be "statistically significant" while being practically meaningless.
Errors	Framework explicitly accounts for Type I error (α) and Type II error (β).	Type I error: rejecting H₀ when it is actually true (false alarm). Type II error: failing to reject H₀ when it is actually false (missed detection). You cannot reduce both simultaneously without increasing n.

✦ KEY TAKEAWAY

Think of the significance level α as a thermostat setting. Setting it low (like 0.01) means you need very strong evidence to "trigger" a rejection of H₀ — you'll rarely sound a false alarm, but you might miss real effects. Setting it higher (like 0.10) makes the test more sensitive but increases the chance of a false alarm. There's no free lunch: every choice of α involves a trade-off between Type I and Type II error.

SECTION 8

Connection to Advanced Concepts

The one-sample t-test and z-interval you have learned are just the starting point of a much broader family of inference methods. As you advance, you will encounter tests that compare two means, test proportions, examine relationships between variables, and more. The table below shows how what you've learned connects to these extensions.

How one-sample inference extends to more advanced testing scenarios.

What You Know Now	Extension	When It's Used
One-sample t-test (H₀: μ = μ₀)	Two-sample t-test	Comparing means of two independent groups (e.g., do boys and girls differ in resting heart rate?)
CI for a mean	CI for a proportion	Estimating the true proportion of a population with a given characteristic (e.g., % of voters supporting a policy)
t-test (numerical data)	Chi-squared (χ²) test	Testing independence or goodness of fit for categorical data (AHL 4.11)
Testing one parameter	ANOVA (Analysis of Variance)	Comparing means across three or more groups simultaneously

The fundamental logic remains identical across all these methods: you state hypotheses, collect data, compute a test statistic and p-value using technology, and compare the p-value to α. Mastering this core process now will make every future test feel like a natural extension rather than something entirely new. In the IB HL course, you will also explore the relationship between confidence intervals and hypothesis tests more deeply, seeing that they are two different lenses on the same underlying statistical reasoning.

SECTION 9

Practice Problems

PROBLEM 1 — CONCEPTUAL

Explain in your own words what it means to say "we are 95% confident" when reporting a 95% confidence interval. Why is it incorrect to say "there is a 95% probability that the true mean lies within this interval"?

PROBLEM 2 — BASIC CALCULATION

A random sample of 40 light bulbs from a factory has a mean lifetime of x̄ = 1020 hours with a sample standard deviation of s = 80 hours. Use your GDC to construct a 95% confidence interval for the true mean lifetime. Report the interval to one decimal place.

PROBLEM 3 — INTERMEDIATE

A coffee chain claims that its medium cups contain a mean of 350 mL of coffee. A customer suspects that the cups are being under-filled. She measures 25 random cups and finds x̄ = 344 mL with s = 12 mL. Perform a one-tailed hypothesis test at the 5% significance level. State H₀ and H₁, report the p-value from your GDC, and write a conclusion in context.

PROBLEM 4 — APPLIED

An environmental agency tests whether the mean dissolved oxygen level in a lake is different from the healthy standard of 8.5 mg/L. They collect a sample of n = 50 water specimens and find x̄ = 8.1 mg/L and s = 1.3 mg/L. (a) Perform a two-tailed t-test at α = 0.05. (b) Construct a 95% confidence interval. (c) Verify that both methods give the same conclusion and explain why.

PROBLEM 5 — CRITICAL THINKING

A student performs a two-tailed t-test at α = 0.05 and obtains p = 0.048. She concludes "the result is statistically significant, so the effect is large and important." Identify two problems with her reasoning, and explain what additional information she would need to properly assess the practical importance of her finding.

SUMMARY

Lesson Summary

A confidence interval provides a range of plausible values for a population parameter, computed as x̄ ± (critical value) × (s / √n). The confidence level (typically 90%, 95%, or 99%) describes the long-run reliability of the method. Hypothesis testing follows a five-step process: state H₀ and H₁, choose a significance level α, use your GDC to find the p-value, and compare it to α. If p < α, reject H₀ and conclude there is sufficient evidence to support H₁; otherwise, do not reject H₀.

For a two-tailed test at level α, the confidence interval and hypothesis test always agree: rejecting H₀ corresponds to μ₀ falling outside the (1 − α) CI. Use a t-test when σ is unknown (almost always in practice) and let your GDC handle the computations. Always write your conclusion in context, and remember that statistical significance does not imply practical importance.