📚 Further Hypothesis Tests | 进阶假设检验
In A-Level Mathematics, hypothesis testing moves beyond simple binomial probabilities. Further hypothesis tests equip you with the tools to handle Poisson means, normal means (with known variance), proportions in large samples, and correlation coefficients. Mastery of these tests demands a solid grasp of distributions, standard errors, and the formal language of significance, p-values, and errors. This article unpacks each test for the Edexcel specification, pairing worked logic with bilingual clarity.
在A-Level数学中，假设检验不再局限于简单的二项分布。进阶假设检验让你掌握处理泊松均值、正态均值（方差已知）、大样本比例以及相关系数的工具。要掌握这些检验，必须透彻理解分布、标准误以及显著性、p值和两类错误的规范语言。本文针对Edexcel考试委员会要求，逐一剖析每种检验，结合示例逻辑与中英双语清晰讲解。
1. Revisiting the Hypothesis Testing Framework | 重温假设检验框架
Every hypothesis test follows a common structure: state the null hypothesis H₀ and the alternative hypothesis H₁, choose a significance level α (often 0.05), compute a test statistic from the sample, and compare it to a critical value or calculate a p-value. A test may be one‑tailed or two‑tailed depending on the direction of the suspected effect.
任何假设检验都遵循统一框架：写出原假设 H₀ 和备择假设 H₁，选定显著性水平 α（通常为0.05），由样本计算检验统计量，再将其与临界值比较或计算p值。依据怀疑的效应方向，检验可以是单尾或双尾。
2. Hypothesis Test for a Poisson Mean | 泊松均值的假设检验
Suppose X ~ he evidence. Because the Poisson distribution is discrete, exact p-values are found by summing probabilities from Poisson tables or software.
设 X ~ Po(λ)，我们希望检验 H₀: λ = λ₀ 对备择 H₁: λ > λ₀、λ < λ₀ 或 λ ≠ λ₀。观测到的总数 ∑x 或平均发生率提供证据。由于泊松分布是离散的，我们需要从泊松分布表或软件中累加概率得到精确p值。
For a one‑tailed test with H₁: λ > λ₀, if the observed count is x, the p‑value = P(X ≥ x | on N(λ₀, λ₀) may be used with a continuity correction, but exact testing is preferred at A‑Level.
对于备择假设 H₁: λ > λ₀ 的单尾检验，若观测计数为x，则p值 = P(X ≥ x | λ = λ₀)。当 p ≤ α 时拒绝 H₀。对于较大的λ₀，可以使用正态近似 N(λ₀, λ₀) 并辅以连续性校正，但在A‑Level中优先使用精确检验。
3. The Normal Mean Test (Known Variance) | 正态均值的检验（方差已知）
normally distributed with known variance σ², testing H₀: μ = μ₀ draws on the sample mean x̄. The test statistic is
当总体服从正态分布且方差σ²已知时，检验 H₀: μ = μ₀ 依赖样本均值 x̄。检验统计量为
Z = (x̄ − μ₀) / (σ / √n)
mal distribution N(0,1) under H₀. Critical values ±zₐ obtain from normal tables. For a two‑tailed test at 5%, the critical region is |Z| > 1.96.
在 H₀ 下该统计量服从标准正态分布 N(0,1)。单双尾临界值 ±zₐ 查正态分布表可得。对于5%显著性水平的双尾检验，临界区域为 |Z| > 1.96。
tral Limit Theorem justifies the same procedure. In Edexcel exam settings, you are normally told that the population is normal or that a large sample renders the assumption safe.
若总体非正态但样本量足够大（n ≥ 30），中心极限定理保证可沿用相同步骤。在Edexcel的考试情境中，通常会明确告知总体服从正态，或指出大样本使得此假设成立。
4. Large Sample Test for a Proportion | 大样本比例检验
To test a population proportion p using H₀: p = p₀, the sample proportion p̂ = x / n is approximately normal when np₀ ≥ 5 and n(1−p₀) ≥ 5. The test statistic is
若要检验总体比例 p，原假设 H₀: p = p₀，当 np₀ ≥ 5 且 n(1−p₀) ≥ 5 时，样本比例 p̂ = x / n 近似服从正态分布。检验统计量为
Z = (p̂ − p₀) / √(p₀(1−p₀) / n)
This Z‑score is compared to normal critical values, exactly as in the mean test. This technique extends the binomial test to larger samples, avoiding cumbersome exact binomial calculations.
将该Z值与正态临界值比较，方法与均值检验完全一致。这一技术将二项检验推广至大样本，避免了繁琐的精确二项计算。
5. Understanding p‑Values | 理解p值
A p‑ ll. If p ≤ α, the result is significant; we reject H₀. If p > α, we do not reject H₀. Smaller p‑values indicate stronger evidence.
p值是在 H₀ 成立的前提下，获得至少与观测结果一样极端的结果的概率。它衡量反对原假设的证据强度。若 p ≤ α，结果显著，我们拒绝 H₀；若 p > α，则不拒绝 H₀。p值越小，证据越强。
In the Edexcel formula booklet, exact p‑values for binomial and Poisson tests can be found using cumulative tables. For normal‑based tests, the p‑value equals 1 − Φ(z) for a right‑tailed test or 2(1 − Φ(|z|)) for a two‑tailed test, where Φ is the standard normal CDF.
在Edexcel公式手册中，二项和泊松检验的精确p值可以通过累积概率表求得。基于正态的检验，p值对右尾检验为 1 − Φ(z)，对双尾检验为 2(1 − Φ(|z|))，其中 Φ 为标准正态累积分布函数。
6. Type I and Type II Errors | 第一类错误和第二类错误
A Type I error occurs when we reject a true H₀; its probability is exactly α. A Type II error happens when we fail to reject a false H₀; its probability is denoted β. The power of a test is 1 − β, the probability of correctly rejecting a false null.
第一类错误是在 H₀ 为真时却拒绝了它，其概率恰为 α。第二类错误是在 H₀ 为假时未能拒绝它，其概率记作 β。检验的功效为 1 − β，即正确拒绝错误原假设的概率。
| Decision \ Truth | H₀ true | H₀ false |
|---|---|---|
| Reject H₀ | Type I error (α) | Correct (1−β) |
| Not reject H₀ | Correct (1−α) | Type II error (β) |
7. Testing for Zero Correlation | 零相关性的检验
ear relationship. To test H₀: ρ = 0 (no correlation) against H₁: ρ ≠ 0, the test statistic
评估线性关联时，积矩相关系数r衡量线性关系的强度。为检验 H₀: ρ = 0（无相关性）对 H₁: ρ ≠ 0，检验统计量为
t = r √n − 2 / √1 − r²
follows a t‑distribution with (n − 2) degrees of freedom. Critical values from the t‑table are compared with the computed t. If |t| exceeds the critical value, we reject H₀ and conclude there is evidence of linear correlation.
服从自由度为 (n − 2) 的t分布。将计算出的t值与t分布表的临界值比较。如果 |t| 超过临界值，则拒绝 H₀，认为有线性相关关系的证据。
Edexcel also provides a table of critical values for r directly, avoiding the t transformation. At A‑Level it is common to use the direct r table; the test statistic approach is included to reinforce conceptual understanding.
Edexcel同时提供直接针对r的临界值表，避免t转换。A‑Level考试中常直接查r表；此处纳入检验统计量是为了强化概念理解。
8. One‑Tailed vs Two‑Tailed Tests | 单尾与双尾检验
A one‑tailed test discerns direction (greater than or less than), placing all of α in one tail. A two‑tailed test splits α equally between both tails, detecting any difference regardless of direction. The choice must be justified by the investigation's aim and specified before seeing the data.
单尾检验判断方向（大于或小于），将全部的α置于一侧尾部。双尾检验则将α平分到两侧尾部，能检测任何方向的差异。必须在审视数据前，根据研究目的确定并明确写出选用单尾或双尾。
A common mistake is to observe a large difference and then switch to a one‑tailed test to achieve significance. This inflates the Type I error and is unacceptable in proper statistical practice.
常见错误是看到一个较大的差异后，再转为单尾检验以达显著。这会放大第一类错误，在规范的统计实践中不可接受。
9. Critical Regions and Power | 临界区域与检验功效
The critical region (or rejection region) is the set of values of the test statistic that leads to rejection of H₀. Its boundaries are determined by the significance level α. The power of a test depends on the true parameter value, sample size, and α. Increasing sample size or α (with caution) raises power, reducing β.
临界区域（或称拒绝域）是会导致拒绝H₀的检验统计量的取值集合，其边界由显著性水平α决定。检验的功效取决于参数真值、样本量和α。增加样本量或谨慎地增大α可提高功效，降低β。
In Edexcel problems, you may be asked to find the critical value for a Poisson or binomial test from tables, or to state the critical region in terms of the sample statistic. Understanding the link between critical region and power helps to interpret non‑significant results.
在Edexcel问题中，你可能被要求通过查表找出泊松或二项检验的临界值，或用样本统计量表达临界区域。理解临界区域与功效的联系有助于解读不显著的结果。
10. Choosing the Right Test and Common Pitfalls | 选择合适的检验及常见误区
Selecting the appropriate test depends on the nature of the data and the hypotheses. Use the Poisson test for counts per unit interval, the normal mean test for continuous symmetric data with known σ, the proportion test for categorical success/failure data, and the correlation test for paired continuous measurements. Always check assumptions: normality, independence, and adequacy of sample size.
选择正确的检验取决于数据性质和假设。单位区间计数用泊松检验；连续对称且σ已知的数据用正态均值检验；分类成功/失败数据用比例检验；配对连续测量用相关性检验。务必检查假设：正态性、独立性和样本量的充分性。
Common pitfalls include confusing p‑value with the probability that H₀ is true, applying a z‑test without checking variance knowledge, and ignoring the effect of sample size on significance. Remember: a small p‑value indicates low compatibility of the data with H₀, not that H₀ is false with a certain probability.
常见误区包括将p值与H₀为真的概率混淆、未确认方差已知就使用z检验、以及忽视样本量对显著性的影响。请记住：小的p值表示数据与H₀的兼容性低，而非H₀以一定概率为假。
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