📚 A-Level Further Mathematics: The Basic Steps of Hypothesis Testing | A-Level进阶数学：假设检验的基本步骤
Hypothesis testing is one of the most powerful and frequently examined tools in A-Level Further Mathematics. It provides a formal framework for making decisions about a population parameter based on sample data, balancing the risk of error against the strength of evidence. Mastering the basic steps allows you to approach every test — whether binomial, Poisson, normal, or t-distribution — with confidence and clarity.
假设检验是A-Level进阶数学中最强大且最常考查的工具之一。它提供了一个正式的框架，使我们能够根据样本数据对总体参数作出决策，在错误风险与证据强度之间取得平衡。掌握基本步骤，你就能自信而清晰地应对每一种检验——无论是二项分布、泊松分布、正态分布还是t分布。
1. What is Hypothesis Testing? | 什么是假设检验？
In statistics, a hypothesis is a claim or statement about a population parameter, such as the mean μ, the proportion p, or the variance σ². Hypothesis testing is the process of using sample data to evaluate how plausible a stated hypothesis is. The underlying logic is a proof by contradiction: we assume a claim is true, then check whether the observed data are unusually unlikely under that assumption. If they are, we reject the claim.
在统计学中，假设是关于总体参数的陈述或主张，例如均值μ、比例p或方差σ²。假设检验是利用样本数据来评估某个既定假设是否可信的过程。其底层逻辑类似于反证法：我们先假定某个主张成立，然后检验在该假设下观测数据是否异常地不可能出现。如果确实如此，我们就拒绝这个主张。
For example, a factory claims that its light bulbs last 1000 hours on average. A sample of bulbs yields a mean of 950 hours. Is this difference due to random chance, or is the factory’s claim false? Hypothesis testing gives a rigorous answer to this question.
例如，某工厂声称其灯泡平均寿命为1000小时。从样本中测得的平均寿命为950小时。这个差异是随机波动造成的，还是工厂的说法是假的？假设检验为这个问题提供了严谨的答案。
2. The Null and Alternative Hypotheses | 原假设与备择假设
Every hypothesis test begins with two competing statements. The null hypothesis, written H₀, is the default assumption that there is no change, no effect, or no difference. It always contains an equals sign, such as μ = 50 or p = 0.3. The alternative hypothesis, written H₁ (or sometimes Hₐ), is what you are trying to find evidence for. It can be one-sided (μ < 50 or μ > 50) or two-sided (μ ≠ 50).
每个假设检验始于两个相互竞争的主张。原假设记作H₀，是默认的假设，即没有变化、没有效应或没有差异。它总是包含等号，如μ = 50或p = 0.3。备择假设记作H₁（有时也记作Hₐ），是你试图寻找证据支持的主张。它可以是单侧的（μ < 50或μ > 50），也可以是双侧的（μ ≠ 50）。
Key rules for writing hypotheses:
- H₀ always includes equality: = , ≥ , or ≤ .
- H₁ never includes equality; it uses < , > , or ≠ .
- Both hypotheses are statements about the population parameter, not the sample statistic.
书写假设的关键规则：
- H₀总是包含等号：=、≥或≤。
- H₁从不包含等号；它使用<、>或≠。
- 两个假设都是关于总体参数的陈述，而不是关于样本统计量的陈述。
In many exam problems, the wording of the question tells you which alternative to use. Words like “greater than” or “improved” suggest a one-tailed test in the positive direction; words like “changed” or “different” suggest a two-tailed test.
在许多考试题目中，问题的措辞会告诉你该使用哪种备择假设。像“大于”或“改进”这类词语暗示正方向的单尾检验；而“改变”或“不同”这类词语则暗示双尾检验。
3. Significance Levels and Critical Regions | 显著性水平与临界域
The significance level, denoted by α, is the probability of rejecting H₀ when H₀ is actually true. This is called a Type I error. Common significance levels in A-Level Further Mathematics are 10%, 5%, and 1%. The significance level determines the critical region: the set of values of the test statistic that would lead us to reject H₀.
显著性水平记为α，是当H₀实际上为真时拒绝H₀的概率。这被称为第一类错误。在A-Level进阶数学中，常见的显著性水平为10%、5%和1%。显著性水平决定了临界域：即检验统计量会导致我们拒绝H₀的取值集合。
The critical value is the boundary between the critical region and the acceptance region. For a two-tailed test at the 5% level, the total probability in the two tails is 0.05, so each tail contains 0.025. For a one-tailed test at the 5% level, the entire 0.05 is placed in one tail.
临界值是临界域与接受域之间的分界点。对于5%水平的双尾检验，两个尾部的总概率为0.05，因此每个尾部各占0.025。对于5%水平的单尾检验，全部0.05放在一个尾部。
P(Reject H₀ | H₀ true) = α
A related concept is the p-value, defined as the probability, assuming H₀ is true, of obtaining a test statistic at least as extreme as the one observed. If the p-value is less than α, we reject H₀; otherwise, we do not reject H₀.
一个相关的概念是p值，其定义为在假设H₀为真的前提下，得到与观测值同等或更极端检验统计量的概率。如果p值小于α，我们拒绝H₀；否则，我们不拒绝H₀。
4. Test Statistics and Their Distributions | 检验统计量及其分布
A test statistic is a quantity computed from the sample data that summarises the evidence against H₀. Its probability distribution under H₀ must be known (or well approximated) so that we can calculate critical values and p-values.
检验统计量是根据样本数据计算出的一个量，用于概括反对H₀的证据。它在H₀成立下的概率分布必须是已知的（或能被很好地近似），这样我们才能计算临界值和p值。
In A-Level Further Mathematics, the most common test statistics arise from the following settings:
在A-Level进阶数学中，最常见的检验统计量来自以下情境：
- Binomial test: X ~ B(n, p₀), used for testing a proportion. Test statistic is X itself.
- Poisson test: X ~ Po(λ₀), used for testing a rate. Test statistic is X itself.
- Normal test (z-test): Z = (x̄ − μ₀) / (σ / √n), used when σ is known.
- t-test: T = (x̄ − μ₀) / (s / √n), used when σ is unknown and estimated by the sample standard deviation s.
二项检验： X ~ B(n, p₀)，用于检验比例。检验统计量为X本身。
泊松检验： X ~ Po(λ₀)，用于检验速率。检验统计量为X本身。
正态检验（z检验）： Z = (x̄ − μ₀) / (σ / √n)，在σ已知时使用。
t检验： T = (x̄ − μ₀) / (s / √n)，在σ未知、用样本标准差s估计时使用。
Notice that for the z-test, the statistic measures how many standard errors the sample mean lies from the hypothesised mean. A large absolute value of Z means the data are far from what H₀ predicts.
注意，对于z检验，该统计量衡量的是样本均值与假设均值之间相差多少个标准误。Z的绝对值越大，说明数据距离H₀的预测越远。
5. One-Tailed and Two-Tailed Tests | 单尾检验与双尾检验
The alternative hypothesis determines the nature of the test. A one-tailed test is used when we only care about deviations in one direction. For example, testing whether a new drug increases recovery rate is a one-tailed test in the positive direction: H₁: p > p₀. The critical region lies entirely in the upper tail of the distribution.
备择假设决定了检验的性质。当我们只关心一个方向上的偏差时，使用单尾检验。例如，检验新药是否能提高治愈率就是一个正方向的单尾检验：H₁: p > p₀。此时临界域完全位于分布的上尾部。
A two-tailed test is used when deviations in either direction are of interest. For example, testing whether the mean height of a population differs from 170 cm uses H₁: μ ≠ 170. The critical region is split between the two tails, with each tail containing α/2 of the probability.
当我们关心两个方向的偏差时，使用双尾检验。例如，检验一个总体的平均身高是否与170厘米不同，使用H₁: μ ≠ 170。此时临界域被分成两个尾部，每个尾部包含α/2的概率。
One-tailed: H₁: μ > μ₀ or H₁: μ < μ₀
Two-tailed: H₁: μ ≠ μ₀
A common mistake is choosing the wrong tail when the test is one-tailed. Always read the question: if it asks whether a value has “increased”, the critical region is in the upper tail; if it asks whether a value has “decreased”, the critical region is in the lower tail.
一个常见错误是在单尾检验中选择错误的尾部。务必仔细阅读题目：如果题目问某个值是否“增加”，临界域在上尾部；如果问是否“减少”，临界域在下尾部。
6. The Step-by-Step Procedure | 分步程序
Although different tests use different statistics, the underlying procedure is always the same. Here are the seven essential steps that apply to any hypothesis test in A-Level Further Mathematics:
尽管不同的检验使用不同的统计量，但其基本程序始终相同。以下是适用于A-Level进阶数学中任何假设检验的七个关键步骤：
Step 1: Define the parameter of interest and state the hypotheses.
Write H₀ and H₁ clearly using symbols. For example, H₀: p = 0.4, H₁: p > 0.4.
第一步：定义感兴趣的参数并写出假设。
用符号清晰地写出H₀和H₁。例如，H₀: p = 0.4，H₁: p > 0.4。
Step 2: Choose the significance level.
The significance level α is usually given in the question, such as 5% or 1%. If not, use 5% as the default.
第二步：选择显著性水平。
显著性水平α通常在题目中给出，如5%或1%。如果没有给出，默认使用5%。
Step 3: State the distribution of the test statistic under H₀.
For example, X ~ B(20, 0.4) or Z ~ N(0, 1). This is essential for calculating probabilities.
第三步：写出H₀成立时检验统计量的分布。
例如，X ~ B(20, 0.4)或Z ~ N(0, 1)。这是计算概率的基础。
Step 4: Compute the observed value of the test statistic from sample data.
Use the appropriate formula. For a binomial test, simply note the observed count x. For a z-test, compute Z using the sample mean.
第四步：根据样本数据计算检验统计量的观测值。
使用相应的公式。对于二项检验，只需记录观测计数x。对于z检验，使用样本均值计算Z。
Step 5: Find the critical region or the p-value.
Use statistical tables or a calculator to determine the critical value(s), or compute the p-value directly.
第五步：确定临界域或p值。
使用统计表或计算器确定临界值，或直接计算p值。
Step 6: Make a decision.
If the observed value lies in the critical region, or if the p-value < α, reject H₀. Otherwise, do not reject H₀.
第六步：作出决策。
如果观测值位于临界域内，或者p值 < α，则拒绝H₀。否则，不拒绝H₀。
Step 7: Write a conclusion in the context of the question.
Do not simply write “reject H₀”. State what this means in words. For example: “There is evidence at the 5% level that the proportion of defective items has increased.”
第七步：结合题目背景写出结论。
不要只写“拒绝H₀”。要用文字说明其含义。例如：“在5%显著性水平下有证据表明次品比例有所上升。”
7. Worked Example: Binomial Test | 实例：二项检验
A coin is suspected of being biased towards heads. It is tossed 20 times, and 14 heads are observed. Test at the 5% significance level whether the coin is biased towards heads.
一枚硬币被怀疑偏向正面。将其抛掷20次，观察到14次正面。在5%显著性水平下检验该硬币是否偏向正面。
Solution:
解法：
Step 1: Let p be the probability of heads. The hypotheses are:
第一步：设p为出现正面的概率。假设为：
H₀: p = 0.5, H₁: p > 0.5 (one-tailed)
Step 2: Significance level α = 0.05.
第二步：显著性水平α = 0.05。
Step 3: Under H₀, X ~ B(20, 0.5), where X is the number of heads.
第三步：在H₀下，X ~ B(20, 0.5)，其中X为正面次数。
Step 4: Observed value is x = 14.
第四步：观测值为x = 14。
Step 5: Since H₁ is p > 0.5, the critical region is in the upper tail. We need the smallest value c such that P(X ≥ c) ≤ 0.05. Using binomial tables or a calculator:
第五步：由于H₁为p > 0.5，临界域在上尾部。我们需要满足P(X ≥ c) ≤ 0.05的最小值c。使用二项分布表或计算器：
- P(X ≥ 15) = 1 − P(X ≤ 14) = 1 − 0.9793 = 0.0207
- P(X ≥ 14) = 1 − P(X ≤ 13) = 1 − 0.9423 = 0.0577
Since 0.0207 < 0.05 but 0.0577 > 0.05, the critical region is X ≥ 15. The observed value x = 14 is not in the critical region.
由于0.0207 < 0.05，而0.0577 > 0.05，因此临界域为X ≥ 15。观测值x = 14不在临界域内。
Alternatively, the p-value is P(X ≥ 14) = 0.0577. Since 0.0577 > 0.05, we do not reject H₀.
或者，p值为P(X ≥ 14) = 0.0577。由于0.0577 > 0.05，我们不拒绝H₀。
Step 6/7: There is insufficient evidence at the 5% significance level to conclude that the coin is biased towards heads.
第六/七步：在5%显著性水平下，没有足够证据表明该硬币偏向正面。
8. Worked Example: Normal (z) Test | 实例：正态（z）检验
A production line fills packets with a mean weight of 500 g and a known standard deviation of 10 g. A sample of 25 packets is found to have a mean weight of 505.5 g. Test, at the 1% significance level, whether the mean weight has increased.
一条生产线灌装的包装袋平均重量为500克，已知标准差为10克。现抽取25个包装袋，样本平均重量为505.5克。在1%显著性水平下检验平均重量是否有所增加。
Solution:
解法：
Step 1: Let μ be the true mean weight. The hypotheses are:
第一步：设μ为真实平均重量。假设为：
H₀: μ = 500, H₁: μ > 500 (one-tailed)
Step 2: Significance level α = 0.01.
第二步：显著性水平α = 0.01。
Step 3: Under H₀, the sample mean x̄ is normally distributed: x̄ ~ N(500, 10²/25), so the test statistic is Z ~ N(0, 1).
第三步：在H₀下，样本均值x̄服从正态分布：x̄ ~ N(500, 10²/25)，因此检验统计量Z ~ N(0, 1)。
Step 4: Compute the test statistic:
第四步：计算检验统计量：
Z = (x̄ − μ₀) / (σ / √n) = (505.5 − 500) / (10 / √25) = 5.5 / 2 = 2.75
Step 5: For a one-tailed test at the 1% level, the critical value is z = 2.3263 (from standard normal tables). The critical region is Z > 2.3263. Since 2.75 > 2.3263, the observed value lies in the critical region.
第五步：对于1%水平的单尾检验，临界值为z = 2.3263（查标准正态表）。临界域为Z > 2.3263。由于2.75 > 2.3263，观测值位于临界域内。
Alternatively, the p-value is P(Z ≥ 2.75) = 0.0030. Since 0.0030 < 0.01, we reject H₀.
或者，p值为P(Z ≥ 2.75) = 0.0030。由于0.0030 < 0.01，我们拒绝H₀。
Step 6/7: There is strong evidence at the 1% significance level that the mean weight of packets has increased.
第六/七步：在1%显著性水平下有很强的证据表明包装袋的平均重量有所增加。
9. Common Mistakes and Exam Tips | 常见错误与考试技巧
Hypothesis testing questions are often worth many marks, and students frequently lose marks through small but avoidable errors. Here are the most common pitfalls:
假设检验题目通常分值较高，学生经常因为一些微小但可避免的错误而失分。以下是最常见的陷阱：
- Writing H₁ with an equals sign. This is always wrong. H₁ must use <, >, or ≠.
- Writing hypotheses about the sample statistic. Hypotheses are about the population parameter, so write μ or p, not x̄ or p̂.
- Using the wrong tail. For H₁: p > p₀, use the upper tail; for H₁: p < p₀, use the lower tail.
- Forgetting to state the distribution under H₀. Always write X ~ B(n, p₀) or Z ~ N(0, 1) before computing probabilities.
- Mixing up “do not reject H₀” with “accept H₀”. We never prove H₀ true; we only say there is insufficient evidence to reject it.
- Writing a conclusion without context. A conclusion such as “reject H₀” alone earns few marks; always relate it to the original problem.
把H₁写成包含等号的形式。这总是错误的。H₁必须使用<、>或≠。
假设写的是样本统计量。假设应针对总体参数，所以写μ或p，而不是x̄或p̂。
使用错误的尾部。对于H₁: p > p₀，使用上尾部；对于H₁: p < p₀，使用下尾部。
忘记说明H₀下的分布。在计算概率之前，务必写出X ~ B(n, p₀)或Z ~ N(0, 1)。
混淆“不拒绝H₀”与“接受H₀”。我们从不证明H₀为真；我们只是说没有足够证据拒绝它。
结论不结合背景。只写“拒绝H₀”得分很少；务必将其与原始问题联系起来。
Another useful tip: when using the “critical region” method with discrete distributions, always show at least two probability calculations to justify your choice of the critical value. Examiners reward clear working.
另一个有用的技巧：在使用离散分布的“临界域”方法时，至少要展示两个概率计算来证明你选择的临界值是合理的。考官会奖励清晰的解题过程。
10. Summary: The Seven-Step Framework | 总结：七步框架
Once you internalise the seven-step framework, every hypothesis test becomes a routine exercise. Let us recap the structure:
一旦你内化了七步框架，每个假设检验都会变成常规练习。让我们回顾一下这个结构：
| Step | 步骤 | Action | 操作 |
|---|---|
| 1 | State H₀ and H₁ | 写出H₀和H₁ |
| 2 | Choose significance level α | 选择显著性水平α |
| 3 | State the distribution under H₀ | 写出H₀下的分布 |
| 4 | Calculate the observed test statistic | 计算观测检验统计量 |
| 5 | Find critical region or p-value | 求出临界域或p值 |
| 6 | Reject or do not reject H₀ | 拒绝或不拒绝H₀ |
| 7 | Write a contextual conclusion | 写出结合背景的结论 |
Remember the golden rule: hypothesis testing is about evidence, not certainty. A significant result means the data are unlikely under H₀, not that H₀ is impossible. Conversely, a non-significant result means the data do not provide strong evidence against H₀, not that H₀ is true.
记住黄金法则：假设检验关乎证据，而非确定性。显著的结果意味着数据在H₀下不太可能出现，但并非H₀不可能。反过来，不显著的结果意味着数据没有提供反对H₀的有力证据，但并非H₀就是真实的。
With practice, you will recognise the patterns instantly: identify the parameter, choose the correct distribution, compute the statistic, compare with the threshold, and communicate your conclusion clearly. This systematic approach will not only earn full marks in exams but also deepen your understanding of statistical inference as a whole.
通过练习，你将能够瞬间识别模式：确定参数、选择正确的分布、计算统计量、与阈值比较，并清晰地表达结论。这种系统化的方法不仅能让你在考试中拿到满分，还能加深你对整个统计推断的理解。
Published by TutorHao | Further Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导