📚 Testing a Hypothesis | 假设检验
In A-Level Mathematics, hypothesis testing is a fundamental tool in statistics that enables us to make informed decisions based on sample data. By setting up competing statements about a population parameter, we can use probability to determine whether observed results provide sufficient evidence to reject the status quo. This article covers the entire Edexcel hypothesis testing syllabus, from stating null and alternative hypotheses to carrying out tests for binomial, Poisson and normal distributions, and interpreting p-values, critical regions and errors.
在 A-Level 数学中，假设检验是统计学的核心工具，使我们能够根据样本数据做出理性判断。通过对总体参数设定一对相互竞争的陈述，我们可以利用概率来判断观察到的结果是否有足够证据否定预设的立场。本文将覆盖 Edexcel 考试大纲中关于假设检验的全部内容，包括零假设与备择假设的提出、二项分布、泊松分布和正态分布的检验，以及p值、临界区域和两类错误的解读。
1. What Is Hypothesis Testing? | 什么是假设检验？
Hypothesis testing is a statistical procedure used to decide whether there is enough evidence in a sample to infer that a certain condition holds for the entire population. We start with a null hypothesis, H₀, which represents the default assumption, and an alternative hypothesis, H₁, which challenges it. By analysing a test statistic based on the sample, we calculate how likely we would be to observe data at least as extreme as ours if H₀ were true. If this probability is very small, we reject H₀ in favour of H₁.
假设检验是一种统计推断方法，用来判断样本数据是否提供了足够证据，支持关于总体的某种主张。我们首先设定零假设 H₀，它代表默认或无变化的情形；同时提出备择假设 H₁，对 H₀ 提出挑战。通过基于样本构造一个检验统计量，我们计算在 H₀ 为真的条件下，得到至少像实际观测那样极端的数据的概率。如果这个概率非常小，我们就拒绝 H₀，转而接受 H₁。
2. Null and Alternative Hypotheses | 零假设与备择假设
= 50. The alternative hypothesis H₁ is the statement we suspect to be true and can be one‑sided (e.g. p > 0.5 or μ < 50) or two‑ nly find evidence against H₀. If the evidence is insufficient, we simply do not reject H₀.
零假设 H₀ 总是包含等号：对于比例可以设 H₀: p = 0.5，对于均值可以设 H₀: μ = 50。备择假设 H₁ 是我们怀疑可能成立的陈述，可以是单侧的（如 p > 0.5 或 μ < 50），也可以是双侧的（p ≠ 0.5，μ ≠ 50）。关键是我们从不直接“接受”H₁；我们只能寻找反对 H₀ 的证据。如果证据不足，我们仅仅是不拒绝 H₀。
3. Significance Level and Significance | 显著性水平与显著性
(alpha), is the probability of rejecting H₀ when it is actually true. Commonly α = 0.05 (5%) or 0.01 (1% the sampling distribution. If the test statistic falls in the critical region, or if the p‑value is smaller than α, we say the result is statistically significant and reject H₀.
显著性水平记为 α，表示当 H₀ 为真时错误拒绝它的概率。通常 α 取 0.05（5%）或 0.01（1%）。在检验之前，我们设定 α，然后在抽样分布的尾部找出截断面积为 α 的临界值。如果检验统计量落入拒绝域，或者 p 值小于 α，我们就说结果是统计显著的，并拒绝 H₀。
4. Test Statistic and Critical Region | 检验统计量与临界区域
tic is a function of the sample data, such as the number of successes X, the sample mean x̄, or a standardised z‑ c that will lead us to reject H₀. For example, when testing H₀: p = 0.5 against H₁: p < 0.5 using a binomial distribution with n=20 at α=0.05, the critical region might be X ≤ 6.
检验统计量是样本数据的函数，例如成功的次数 X、样本均值 x̄ 或标准化的 z 分数。它在 H₀ 下的抽样分布是已知的（二项、泊松、正态等）。临界区域（拒绝域）是导致我们拒绝 H₀ 的检验统计量所有取值的集合。例如，使用 n=20 的二项分布在 α=0.05 下检验 H₀: p = 0.5 对 H₁: p < 0.5，临界区域可能是 X ≤ 6。
5. p‑value Method | p 值方法
The p‑ of obtaining a test statistic at least as extreme as the one actually observed. For a left‑tailed test, p‑value = P(X ≤ observed); for a right‑tailed test, p‑value = P(X ≥ observed). In a two‑tailed test, we consider both tails and often double the smaller one‑tailed probability, provided the distribution is symmetric. If p‑value < α, we reject H₀.
p 值是在 H₀ 为真的条件下，得到至少与实际观测值同样极端的检验统计量的概率。对于左侧检验，p 值 = P(X ≤ 观测值)；对于右侧检验，p 值 = P(X ≥ 观测值)。在双侧检验中，我们考虑两侧尾部，通常将较小的一侧概率乘以 2（在对称分布下）。如果 p 值 < α，我们拒绝 H₀。
6. One‑tailed and Two‑tailed Tests | 单尾与双尾检验
If the alternative hypothesis specifies a direction (e.g. μ > 20), we use a one‑tailed test and place the entire significance level α in that tail. If the alternative is simply ‘different from’ (μ ≠ 20), we use a two‑tailed test with α/2 in each tail. The choice affects the critical values and the p‑value calculation. Edexcel questions will make the direction clear or ask you to state it based on the context.
如果备择假设指明了方向（如 μ > 20），我们用单尾检验，并将全部显著性水平 α 放在该侧尾部。如果备择假设只是“不等于”（μ ≠ 20），则用双尾检验，每侧尾部分配 α/2。这一选择会影响临界值和 p 值的计算。Edexcel 的考题会明确方向，或要求你根据上下文进行判断。
7. Binomial Hypothesis Testing | 二项假设检验
When a discrete variable X ∼ toss a coin 20 times and obtain 6 heads, testing H₀: p = 0.5 against H₁: p ≠ 0.5 at the 5% level. The p‑ = 2 × 0.0577 ≈ 0.1154. Since this exceeds 0.05, we do not reject H₀. If the test were one‑tailed, e.g. H₁: p < 0.5, p‑value = P(X ≤ 6) = 0.0577, still not significant at 5%.
当使用离散变量 X ∼ B(n, p) 时，我们通过二项分布计算精确概率。假设投掷一枚硬币 20 次得到 6 次正面，在 5% 水平下检验 H₀: p = 0.5 对 H₁: p ≠ 0.5。则 p 值 = 2 × P(X ≤ 6) = 2 × 0.0577 ≈ 0.1154。因为大于 0.05，我们不拒绝 H₀。如果检验是单侧的，例如 H₁: p < 0.5，p 值 = P(X ≤ 6) = 0.0577，在 5% 水平下依然不显著。
8. Hypothesis Testing for a Proportion Using Normal Approximation | 比例假设检验（正态近似）
For large n, the sample proportion p̂ is approximately normal: p̂ ∼ N(p, p(1‑p)/n). We use the test statistic z = (p̂ − p) / √(p(1‑p)/n). Critical values come from the standard normal distribution, e.g. ±1.96 for a two‑tailed test at α = 0.05. This method is valid when np > 5 and n(1‑p) > 5. Continuity corrections can be applied for increased accuracy.
当 n 较大时，样本比例 p̂ 近似服从正态分布：p̂ ∼ N(p, p(1‑p)/n)。我们使用检验统计量 z = (p̂ − p) / √(p(1‑p)/n)。临界值取自标准正态分布，例如 α = 0.05 的双尾检验用 ±1.96。当 np > 5 且 n(1‑p) > 5 时该方法有效。为提高精确度可加入连续性校正。
9. Poisson Hypothesis Testing | 泊松假设检验
For rate data, X ∼ λ > 4 at α = 0.05, we find P(X ≥ 10) = 1 − P(X ≤ 9) ≈ 0.0081. Since 0.0081 < 0.05, we reject H₀ and conclude the rate has increased.
对于计数率数据，X ∼ Po(λ)。假设一个呼叫中心宣称每小时平均呼叫数为 λ = 4。我们监测一个小时并观测到 10 次呼叫。在 α = 0.05 下检验 H₀: λ = 4 对 H₁: λ > 4，计算 P(X ≥ 10) = 1 − P(X ≤ 9) ≈ 0.0081。因为 0.0081 < 0.05，我们拒绝 H₀，并认为呼叫率有所上升。
10. Hypothesis Testing for the Mean of a Normal Distribution | 正态分布均值的假设检验
n and the data are normal (or n ≥ 30 by the Central Limit Theorem), we test μ using z = (x̄ − μ₀) / (σ/√n xample, a chocolate bar’s stated weight is 50 g with σ = 2 g. A sample of 25 bars yields x̄ critical values are ±1.96, so we reject H₀ and conclude the mean weight differs from 50 g.
当总体方差 σ² 已知且数据服从正态分布（或根据中心极限定理 n ≥ 30），我们用 z = (x̄ − μ₀) / (σ/√n) 来检验均值 μ。例如，某巧克力棒标称重量为 50 g，σ = 2 g。抽取 25 根样本测得 x̄ = 48.5 g。在 α = 0.05 下检验 H₀: μ = 50 对 H₁: μ ≠ 50，检验统计量 z = (48.5 − 50) / (2/5) = −3.75。双侧临界值为 ±1.96，故拒绝 H₀，认为平均重量与 50 g 有显著差异。
11. Type I and Type II Errors | 第一类错误与第二类错误
A Type I error occurs when we reject a true null hypothesis; its probability is α, the significance level. A Type II error occurs when we fail to reject a false null hypothesis; its probability is β. The power of a test is 1 − β, the chance of correctly rejecting a false H₀. Increasing sample size or α can raise power, but raising α also increases Type I error risk. A‑Level questions often ask you to describe, but not always calculate, these errors.
第一类错误发生在 H₀ 为真时却被拒绝，其概率等于显著性水平 α。第二类错误发生在 H₀ 为假时却没有被拒绝，其概率记为 β。检验的功效为 1 − β，即正确拒绝一个错误 H₀ 的概率。增大样本量或 α 可以提高功效，但提高 α 也会增加犯第一类错误的风险。A‑Level 题目常要求描述这些错误，而不一定要求计算。
12. Common Pitfalls and Tips for the Exam | 常见错误与应试技巧
Always write the hypotheses clearly, using the parameter in question (p, λ, μ). Do not confuse the sample statistic with the population parameter. When conducting a two‑tailed test, remember to double the one‑tailed p‑value (for symmetric distributions) or compare the test statistic with two critical values. In binomial and Poisson tests, probabilities should be found using the distribution stated under H₀. Finally, draw a conclusion in the context of the problem – “there is sufficient evidence at the 5% level to suggest that…” or “there is insufficient evidence…” – and be precise about whether you ‘reject H₀’ or ‘do not reject H₀’.
务必清晰地写出假设，使用相关的总体参数（p、λ、μ）。不要混淆样本统计量与总体参数。进行双侧检验时，记得将单侧 p 值加倍（对于对称分布）或用两个临界值进行比较。在二项分布和泊松分布检验中，应在 H₀ 所确定的分布下计算概率。最后，结合问题背景给出结论——“在 5% 的显著性水平下，有充分证据表明……”或“没有充分证据……”，并准确表述为“拒绝 H₀”或“不拒绝 H₀”。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图