Mastering Statistical Hypothesis Testing for Proportions in A-Level Maths

Introduction to Hypothesis Testing

In A-Level Statistics, hypothesis testing is a fundamental tool used to make inferences about a population based on a sample. When we deal with proportions, we are typically looking at a binomial scenario where there are two possible outcomes: success or failure. By testing a hypothesis, we determine whether an observed change in proportion is statistically significant or simply the result of random chance.

Understanding this topic is vital for your exams. You will be expected to set up null and alternative hypotheses, identify critical regions, and draw valid conclusions based on a chosen significance level. This guide will walk you through the logic and the mechanics required to excel in these questions.

Defining Hypotheses

Every hypothesis test begins by defining two competing statements about the population parameter, $p$. The null hypothesis ($H_0$) represents the status quo—the assumption that nothing has changed. The alternative hypothesis ($H_1$) represents the claim we are testing.

• $H_0: p = p_0$ (The population proportion remains unchanged)
• $H_1: p < p_0$, $p > p_0$, or $p \neq p_0$ (The proportion has decreased, increased, or changed)

If $H_1$ uses an inequality ($<$ or $>$), it is a one-tailed test. If it uses the inequality ($ eq$), it is a two-tailed test, meaning we are looking for evidence of change in either direction.

The Significance Level and Critical Regions

The significance level ($\alpha$) is the threshold for rejecting $H_0$. Common values are 0.05 (5%) or 0.01 (1%). If the probability of observing our result (or something more extreme) is less than $\alpha$, we reject $H_0$.

A critical region is the set of values for the test statistic that leads to the rejection of $H_0$. For a binomial distribution $X \sim B(n, p)$, we find the critical value $c$ such that $P(X \leq c) \leq \alpha$ (for a lower tail) or $P(X \geq c) \leq \alpha$ (for an upper tail).

Worked Example 1: One-Tailed Test

A shopkeeper claims that 20% of customers buy a specific brand of tea. A new manager suspects this proportion has increased. They sample 20 customers and find 8 bought the tea. Test this at the 5% significance level.

1. Hypotheses: $H_0: p = 0.2$, $H_1: p > 0.2$.
2. Distribution: $X \sim B(20, 0.2)$.
3. Calculation: We want $P(X \geq 8)$. Using binomial tables or a calculator: $P(X \geq 8) = 1 - P(X \leq 7) \approx 1 - 0.9679 = 0.0321$.
4. Conclusion: Since $0.0321 < 0.05$, the result is significant. We reject $H_0$ and conclude there is sufficient evidence to suggest the proportion has increased.

Worked Example 2: Two-Tailed Test

A factory claims 10% of items are defective. A quality inspector takes a sample of 50 items and finds 10 are defective. Test at the 10% significance level.

1. Hypotheses: $H_0: p = 0.1$, $H_1: p \neq 0.1$.
2. Distribution: $X \sim B(50, 0.1)$.
3. Significance: For a 10% two-tailed test, we split the significance into 5% in each tail.
4. Critical Region: Find $c_1$ such that $P(X \leq c_1) \leq 0.05$ and $c_2$ such that $P(X \geq c_2) \leq 0.05$.
   • $P(X \leq 1) = 0.0338$ (less than 0.05), $P(X \leq 2) = 0.1117$ (greater than 0.05). So, lower critical region is $X \leq 1$.
   • $P(X \geq 9) = 1 - P(X \leq 8) = 1 - 0.9421 = 0.0579$ (greater than 0.05), $P(X \geq 10) = 1 - P(X \leq 9) = 1 - 0.9755 = 0.0245$ (less than 0.05). So, upper critical region is $X \geq 10$.
5. Conclusion: Our observed value is 10. Since $10 \geq 10$, it falls in the critical region. We reject $H_0$.

Common Mistakes

• Confusing the tails: Always check if the question implies a one-tailed or two-tailed test. A two-tailed test requires splitting the significance level.
• Incorrect Hypotheses: Ensure your hypotheses are written in terms of the population parameter $p$, not the sample proportion $\hat{p}$.
• Misinterpreting the p-value: Remember that a low p-value means the result is unlikely under $H_0$, which is why we reject it. Do not say we "accept" $H_0$; we simply "fail to reject" it.

FAQ

What is the difference between the significance level and the actual significance level? The significance level is the threshold you choose (e.g., 5%). The actual significance level is the exact probability of falling into the critical region, which is often slightly different due to the discrete nature of the binomial distribution.

When should I use a normal approximation? Use a normal approximation to the binomial distribution only when $n$ is large and $p$ is close to 0.5, typically when $np > 5$ and $nq > 5$.

How do I know if a test is one-tailed or two-tailed? Look for keywords. "Increased" or "decreased" implies one-tailed. "Changed" or "different" implies two-tailed.

Conclusion

Hypothesis testing for proportions is a cornerstone of A-Level Statistics. By mastering the steps of defining hypotheses, calculating probabilities, and comparing them to your significance level, you can tackle any exam question with confidence. For more practice, head over to MathInstructor AI to generate a free, narrated animated lesson on this topic.