Hypothesis Testing for Zero Correlation | 零相关假设检验

📚 Hypothesis Testing for Zero Correlation | 零相关假设检验

Hypothesis testing for zero correlation is a fundamental statistical procedure in A-Level Mathematics. It allows us to determine whether the linear relationship observed in a sample of bivariate data is strong enough to conclude that a genuine correlation exists in the population, or whether it could simply be due to chance. This topic sits at the intersection of correlation analysis and formal inference, making it vital for any student aiming for top marks in Edexcel's applied modules.

零相关的假设检验是A-Level数学中的一项核心统计方法。它可以帮助我们判断在双变量样本数据中观察到的线性关系，是否足以推断总体中存在真正的相关关系，还是仅仅由偶然因素造成。这一主题融合了相关分析与形式化推断，是学生在Edexcel应用模块中夺取高分的关键内容。

1. Introduction to Correlation and Hypothesis Testing | 相关性与假设检验简介

Correlation measures the strength and direction of a linear relationship between two variables, such as x and y. The population correlation coefficient is denoted by the Greek letter ρ (rho). In real-world applications, ρ is almost always unknown, so we estimate it using the sample product moment correlation coefficient r. However, r can vary from sample to sample, which is why we need a formal test to decide whether ρ is likely to be zero based on a given r value and sample size n.

相关性衡量两个变量（例如 x 和 y）之间线性关系的强度和方向。总体相关系数用希腊字母 ρ (rho) 表示。在实际应用中，ρ 几乎总是未知的，因此我们使用样本积矩相关系数 r 来进行估计。然而，r 会因样本不同而波动，所以我们需要一个正式的检验，根据给定的 r 值和样本容量 n，判断 ρ 是否很可能为零。

2. The Null and Alternative Hypotheses for Zero Correlation | 零相关的原假设与备择假设

ive hypothesis depends on the context of the investigation. For a two-tailed test, which checks for any correlation, we write H₁: ρ ≠ 0. For a one-tailed test, we test for positive correlation (H₁: ρ > 0) or negative correlation (H₁: ρ < 0). Choosing the correct tail is crucial: if the question says 'test whether there is positive correlation', use a one-tailed test; if it says 'test whether there is correlation', use a two-tailed test.

原假设总是陈述总体中不存在线性相关：H₀: ρ = 0。备择假设取决于研究的背景。对于检验是否存在任何相关的双尾检验，我们写作 H₁: ρ ≠ 0。对于单尾检验，我们检验正相关 (H₁: ρ > 0) 或负相关 (H₁: ρ < 0)。选择正确的尾部至关重要：如果题目说“检验是否存在正相关”，使用单尾检验；如果说“检验是否存在相关”，则使用双尾检验。

3. Understanding the Sample Correlation Coefficient r | 理解样本相关系数 r

The sample correlation coefficient r is calculated using the formula r = Sxy / √Sxx × Syy x)(Σy)/n, Sxx = Σx² - (Σx)²/n, and Syy = Σy² - (Σy)²/n. The value of r always lies between –1 and 1. An r close to 1 indicates strong positive correlation, an r close to – hip. However, a non-zero r in a sample does not automatically mean ρ ≠ 0 in the population – that is what the hypothesis test decides.

样本相关系数 r 使用公式 r = Sxy / √Sxx × Syy 计算。其中 Sxy = Σ(xy) – (Σx)(Σy)/n，Sxx = Σx² – (Σx)²/n，Syy = Σy² – (Σy)²/n。r 的值总是在 –1 到 1 之间。r 接近 1 表明强正相关，接近 –1 表明强负相关，接近 0 表明几乎没有线性关系。然而，样本中存在非零的 r 并不意味着总体中 ρ ≠ 0——这正是假设检验需要判定的。

4. Steps for a Hypothesis Test (Critical Value Method) | 假设检验步骤（临界值法）

To conduct a hypothesis test for zero correlation using critical values, follow these steps systematically: (1) State H₀ and H₁ clearly, including the direction if one-tailed. (2) Determine the significance level α, often 5% or 1%. (3) Find the critical value from the PMCC table using n, α, and the tail type. (4) Calculate the test statistic r from the sample data. (5) Compare |r| with the critical value, and decide whether to reject H₀. (6) Write a conclusion in the context of the problem, never just 'reject H₀' or 'do not reject H₀'.

使用临界值法进行零相关假设检验时，请按以下步骤操作：(1) 清晰陈述 H₀ 和 H₁，包括单尾的方向。(2) 确定显著性水平 α，通常为 5% 或 1%。(3) 利用 n、α 和尾部类型从 PMCC 表中查找临界值。(4) 根据样本数据计算检验统计量 r。(5) 将 |r| 与临界值进行比较，决定是否拒绝 H₀。(6) 结合问题背景写出结论，绝不能只写“拒绝 H₀”或“不拒绝 H₀”。

5. Using the PMCC Critical Values Table | 使用积矩相关系数临界值表

ample sizes n and significance levels. For example, with n = 10 and a two-tailed test at 5% task is to extract the correct value from a given table extract. Note that the table usually gives values for one-tailed tests, and you must select the appropriate column based on whether the test is one- or two-tailed. Below is a sample extract:

积矩相关系数表给出了不同样本容量 n 和显著性水平下 r 的临界值。例如，当 n = 10 时，在 5% 显著性水平下的双尾检验临界值可能是 0.6319。这意味着，如果你计算得到的 r 的绝对值超过 0.6319，就拒绝 H₀。从给定的表格节选中提取正确数值是常见的考试任务。注意，表格通常给出单尾检验的值，你必须根据检验是单尾还是双尾选择恰当的列。下面是一个示例节选：

n	5% (one-tail)	2.5% (one-tail)	0.5% (one-tail)
8	0.6215	0.7067	0.8343
9	0.5822	0.6664	0.7977
10	0.5494	0.6319	0.7646
11	0.5214	0.6021	0.7348

For a two-tailed test at 5%, you would use the 2.5% one-tail column because 5% total is split equally between the two tails. Understanding this equivalence is key to avoiding common mistakes.

对于 5% 显著性水平的双尾检验，应使用 2.5% 的单尾列，因为总显著性水平 5% 均分在两个尾部。理解这一等价关系是避免常见错误的关键。

6. Example 1: Testing for Positive Correlation (One-tailed) | 例1：检验正相关（单尾）

y correlated. From a random sample of 12 students, she calculates r = 0.487. Test at the 5% significance level whether there is evidence of positive correlation.

广告 / Sponsored

一位研究者相信复习时间 (x) 与考试成绩 (y) 呈正相关。她从一个包含 12 名学生的随机样本中计算出 r = 0.487。在 5% 的显著性水平下检验是否有正相关的证据。

Hypotheses: H₀: ρ = 0, H₁: ρ > 0 (one-tailed). n = 12, α = 5%. From the PMCC table, the one-tailed 5% sult is significant. We reject H₀ and conclude that there is sufficient evidence, at the 5% level, of a positive correlation between revision hours and exam score in the population.

假设：H₀: ρ = 0，H₁: ρ > 0（单尾）。n = 12，α = 5%。根据 PMCC 表，n = 12 时单尾 5% 临界值为 0.4575（假设典型表格）。由于 r = 0.487 > 0.4575，结果显著。我们拒绝 H₀，并得出结论：在 5% 水平上，有充分证据表明总体中复习时间与考试成绩存在正相关。

7. Example 2: Testing for Any Correlation (Two-tailed) | 例2：检验是否存在相关（双尾）

e daily temperature (°C) and the number of hot drinks sold over 9 days. She calculates r = –0.624. Test at the 1% significance level whether there is any correlation between temperature and sales.

一家咖啡馆的老板记录了 9 天中每天的气温（°C）和热饮销量，计算得到 r = –0.624。在 1% 的显著性水平下检验温度与销量之间是否存在任何相关关系。

Hypotheses: H₀: ρ = 0, H₁: ρ ≠ 0 (two-tailed). n = 9, α = 1% overall, so for a two-tailed test we need the 0.5% one-tail column. From the table extract for n = 9, the 0.5% critical value is 0.7977. Compare the absolute value of r: |–0.624| 0.624. Since 0.624 < 0.7977, we do not reject H₀. There is insufficient evidence at the 1% one-tailed test might have been significant, but the two-tailed test is more conservative.

假设：H₀: ρ = 0，H₁: ρ ≠ 0（双尾）。n = 9，总体 α = 1%，因此双尾检验需要使用 0.5% 的单尾列。从 n = 9 的表格节选中，0.5% 临界值为 0.7977。比较 r 的绝对值：|–0.624| = 0.624。因为 0.624 < 0.7977，我们不拒绝 H₀。在 1% 显著性水平下，没有足够证据表明温度与热饮销量之间存在相关。请注意，单尾检验可能显著，但双尾检验更为保守。

8. The p-value Method | p 值法

tion as extreme as, or more extreme than, the observed r, assuming H₀ is true. If the p-value < α, we reject H₀. For a two-tailed test, double the one-tailed p-value. In Edexcel exams, you may be given a p-value and asked to interpret it. For instance, if r = 0.31 with n = 30 and the one-tailed p-value is 0.045, then for a two-tailed test the p-value is 0.09. At α = 0.05, we would not reject H₀ because 0.09 > 0.05. The p-value approach gives a more continuous measure of evidence against H₀.

现代软件通常不将 r 与临界值比较，而是输出一个 p 值。p 值是在 H₀ 成立的条件下，得到与观测 r 同等极端或更极端的样本相关系数的概率。如果 p 值 < α，我们拒绝 H₀。对于双尾检验，将单尾 p 值加倍。在 Edexcel 考试中，你可能会得到一个 p 值并被要求解读。例如，如果 r = 0.31，n = 30，单尾 p 值为 0.045，那么双尾检验的 p 值就是 0.09。当 α = 0.05 时，我们不会拒绝 H₀，因为 0.09 > 0.05。p 值法提供了反对 H₀ 的连续度量。

9. Checking Assumptions | 验证假设条件

For the hypothesis test on ρ to be valid, certain assumptions must be met. The data should be a random sample from the bivariate population. Both variables should follow a bivariate normal distribution, which can be checked informally by plotting a scatter diagram: the points should form an elliptical cloud, and the relationship should appear linear. Outliers can severely distort r, so they must be investigated. Also, the test assumes that the variables are measured on a continuous scale and that the sample size n is at least 3. In many A-Level questions, you are told to assume these conditions hold, but in contextual questions you may need to comment on whether a scatter plot supports linearity.

要使 ρ 的假设检验有效，必须满足某些假设条件。数据应是从双变量总体中抽取的随机样本。两个变量应服从双变量正态分布，可以通过绘制散点图非正式地检查：点应形成一个椭圆形云团，且关系应呈线性。异常值会严重扭曲 r，因此必须加以调查。此外，检验假定变量是在连续尺度上测量的，并且样本容量 n 至少为 3。在许多 A-Level 题目中，会直接告知假设这些条件成立，但在情境题中，你可能需要评论散点图是否支持线性关系。

10. Interpretation and Conclusion | 结果解读与结论

Writing a good conclusion is essential for exam success. A weak conclusion might say 'reject H₀', while a strong one states: 'Since r = 0.68 exceeds the 5% critical value of 0.6021 for n = 11 (two-tailed), there is sufficient evidence at the 5% s a linear correlation between engine size and fuel consumption in the population of cars.' ays refer back to the context, mention the significance level, and be careful not to claim 'cause and effect'. Correlation does not imply causation, and students should avoid language like 'proves' or 'shows that x causes y'.

撰写良好的结论对考试成功至关重要。一个薄弱的结论可能只写“拒绝 H₀”，而一个好的结论则是：“由于 r = 0.68 超过了 n = 11 时 5% 双尾检验临界值 0.6021，因此在 5% 显著性水平下有足够证据表明，在汽车总体中引擎排量与燃油消耗之间存在线性相关。” 务必结合背景，提及显著性水平，并小心不要宣称“因果关系”。相关并不意味着因果，学生应避免使用“证明”或“表明 x 导致 y”这类措辞。

11. Common Pitfalls | 常见错误

Many students lose marks by mixing up one-tailed and two-tailed critical values. Always check whether H₁ uses >, <, or ≠. Another common mistake is forgetting to take the absolute value of r when comparing with the critical value; the critical value itself is always non‑negative. Using the wrong n is also frequent – n is the number of pairs, not the sum of data points. Finally, do not confuse the critical value for correlation with those for the t‑test or normal distribution; they come from different tables. When stating the conclusion, avoid definitive phrases like 'this proves the correlation exists' – hypothesis testing is about evidence, not proof.

许多学生因混淆单尾和双尾临界值而失分。务必检查 H₁ 使用的是 >、< 还是 ≠。另一个常见错误是比较时忘记取 r 的绝对值，临界值本身总是非负的。使用错误的 n 也很常见——n 是数据对的数量，不是数据点的总数。最后，不要将相关系数的临界值与 t 检验或正态分布的临界值混淆，它们来自不同的表格。陈述结论时，避免使用“这证明了相关存在”等绝对化短语——假设检验是关于证据，而非证明。

12. Exam Tips and Summary | 考试技巧与总结

To excel in hypothesis testing for zero correlation, practice extracting critical values quickly from the PMCC table, paying close attention to the tail and significance level. Show all steps clearly in your working: hypotheses, significance level, critical value, comparison, and contextual conclusion. If given a p-value, simply compare it with α for the appropriate tail. Remember that the test only assesses linear correlation; a non-significant result does not rule out a curved relationship. Review past Edexcel papers to get familiar with the phrasing of questions. With methodical preparation, this topic becomes a reliable source of marks.

要在零相关假设检验中取得优异成绩，应练习快速从 PMCC 表中提取临界值，并密切关注尾部和显著性水平。在解答中清晰展示所有步骤：假设、显著性水平、临界值、比较以及情境化结论。如果给出 p 值，只需将其与对应尾部的 α 比较。请记住，该检验仅评估线性相关；不显著的结果并不排除曲线关系的存在。复习过往的 Edexcel 试卷，熟悉题目的措辞。通过有条理的准备，这一主题将成为可靠的得分来源。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）