📚 Type I and Type II Errors | 第一类错误与第二类错误
In A-Level Edexcel Mathematics, hypothesis testing provides a systematic framework for making decisions about population parameters using sample data. However, because we are dealing with uncertainty, there is always a chance that our conclusion is wrong. Type I and Type II errors are the two kinds of mistakes that can occur in a test, and understanding them is crucial for interpreting results in real-world contexts such as medical trials, quality control, and scientific research.
在 A-Level Edexcel 数学中，假设检验提供了一个系统框架，利用样本数据对总体参数作出决策。然而，由于我们面对的是不确定性，结论出错的可能性始终存在。第一类错误和第二类错误是检验中可能发生的两种错误类型，理解它们对于解读医学试验、质量控制和科学研究等实际场景的结果至关重要。
1. The Structure of a Hypothesis Test | 假设检验的结构
Every hypothesis test begins by stating a null hypothesis H₀ and an alternative hypothesis H₁. Based on a test statistic calculated from the sample, we compare the result against a critical region. If the statistic falls inside the critical region, we reject H₀ in favour of H₁; otherwise, we do not reject H₀. This conclusion is not a proof but an inference under a predefined significance level α.
每个假设检验首先需要陈述原假设 H₀ 和备择假设 H₁。根据从样本中计算出的检验统计量，我们将结果与临界域进行对比。如果统计量落入临界域，我们拒绝 H₀ 而支持 H₁；否则，我们不拒绝 H₀。这一结论并非证明，而是在预先设定的显著性水平 α 下的一种推断。
2. Null and Alternative Hypotheses | 原假设与备择假设
The null hypothesis H₀ usually represents the status quo or no effect, such as 'the coin is fair' (p = 0.5) or 'the mean is 100' he choice between one-tailed and two-tailed tests affects the location of the critical region and the potential for errors.
原假设 H₀ 通常代表现状或无效应，例如“硬币是公平的”（p = 0.5）或“均值为 100”。备择假设 H₁ 表明存在效应，以单尾（p > 0.5 或 p < 0.5）或双尾（p ≠ 0.5）的形式陈述。选择单尾检验还是双尾检验会影响临界域的位置以及犯错的可能性。
3. What Is a Type I Error? | 什么是第一类错误？
inical trial, it would be like concluding a new medicine works when it actually does not. The probability of committing a Type I error is exactly the significance level α, so if α = 0.05, there is a 5% risk of rejecting a true H₀.
第一类错误发生在原假设 H₀ 为真，但检验结果却导致我们拒绝它的时候。这通常被称为假阳性。在临床试验的情境中，这就像是在新药实际上无效时却得出了它有效的结论。犯第一类错误的概率恰好就是显著性水平 α，因此若 α = 0.05，拒绝一个真实的 H₀ 的风险就是 5%。
4. What Is a Type II Error? | 什么是第二类错误？
A Type II error happens when the null hypothesis H₀ is false, yet we fail to reject it. This is a false negative. Using the medical example, a Type II error would mean failing to detect that a working drug actually has an effect. The probability of a Type II error is denoted by β, and it depends on the true value of the population parameter, the sample size, and the significance level chosen.
第二类错误发生在原假设 H₀ 为假，但我们却未能拒绝它的时候。这是一种假阴性。继续以医学为例，第二类错误就意味着未能检测出一种实际上有效的药物所具有的效果。第二类错误的概率用 β 表示，它取决于总体参数的真实值、样本量以及所选择的显著性水平。
5. Significance Level and Its Role | 显著性水平及其作用
mmon values are 0.05, 0.01, or 0.10. By setting α, we directly control the critical region' ize. For a one-tailed test with α = 0.05, for instance, the critical region contains the 5% ect H₀, which lowers the risk of a Type I error but increases the risk of a Type II error.
显著性水平 α 是检验者愿意接受的犯第一类错误的最大概率。在 Edexcel 考题中，常见取值为 0.05、0.01 或 0.10。通过设定 α，我们直接控制了临界域的大小。例如，对于 α = 0.05 的单尾检验，临界域包含了 H₀ 下最极端的 5% 结果。降低 α 会使拒绝 H₀ 变得更加困难，这虽然降低了第一类错误的风险，却增加了第二类错误的风险。
6. Probability of Type II Error and Power | 第二类错误的概率与检验功效
The probability of a Type II error, β, is calculated under the assumption that a specific alternative value is true. For a test of proportion, if the true p is different from H₀, we determine β by finding the probability that the test statistic falls outside the critical region for that true p. The power of a test is defined as 1 – β, representing the probability of correctly rejecting a false H₀. High power is desirable, often targeted at 0.8 or higher in research.
第二类错误的概率 β 是在假定某个特定的备择值为真的条件下计算得出的。对于比例检验，如果真实的 p 不同于 H₀，我们可以通过计算检验统计量在该真实 p 下未落入临界域的概率来确定 β。检验的功效定义为 1 – β，它表示正确拒绝一个错误 H₀ 的概率。我们希望功效较高，研究中通常以 0.8 或更高为目标。
7. Binomial Hypothesis Test: Type I and Type II Error Example | 二项分布假设检验：两类错误示例
g a sample of n = 20 patients. The significance level is α = 0.05. The test statistic is X ~ find k = 7, so the critical region is X ≤ 7. If the true cure rate is actually p = 0.4, then β = P(not rejecting H₀ | p = 0.4) = P(X > 7 | p = 0.4) = 1 – P(X ≤ 7 | 0.4) tables, P(X ≤ 7) ≈ 0.4159, so β ≈ 0.5841. The power is roughly 0.4159, which is low.
假设有一种新药声称其治愈疾病的概率 p = 0.6。我们利用容量为 n = 20 的患者样本检验 H₀: p = 0.6 和 H₁: p < 0.6，显著性水平 α = 0.05。在 H₀ 下检验统计量为 X ~ B(20, 0.6)。我们需要找出最大的 k 使得 P(X ≤ k) ≤ 0.05，从而确定临界域。根据二项分布累积表，我们可能得到 k = 7，因此临界域为 X ≤ 7。若真实的治愈率实际上为 p = 0.4，则 β = P(未拒绝 H₀ | p = 0.4) = P(X > 7 | p = 0.4) = 1 – P(X ≤ 7 | p = 0.4)。查阅 B(20, 0.4) 表，P(X ≤ 7) ≈ 0.4159，故 β ≈ 0.5841。检验功效约为 0.4159，属于较低水平。
8. Normal Distribution Context: Mean Testing | 正态分布背景：均值检验
r H₀: μ = 50 against H₁: μ > 50, with σ = 6 and a sample of size n = 36, the sample mean X̄ ~ 05, the critical value is 50 + 1.6449 = 51.6449. If the true mean μ = 52, then under H₁, X̄ ~ N(52, 1). The probability of Type II error is β = P(X̄ < 51.6449 | μ = 52) = P(Z < (51.6449 – 52)/1) = P(Z < –0.3551) ≈ 0.361. So the power is 0.639.
在检验总体均值时，我们通常假定方差已知且服从正态分布。对于 H₀: μ = 50 和 H₁: μ > 50，已知 σ = 6，样本容量 n = 36，则在 H₀ 下样本均值 X̄ ~ N(50, 6²/36)，即 N(50, 1)。在 α = 0.05 下，临界值为 50 + 1.6449 = 51.6449。若真实均值 μ = 52，则在 H₁ 下 X̄ ~ N(52, 1)。第二类错误的概率为 β = P(X̄ < 51.6449 | μ = 52) = P(Z < (51.6449 – 52)/1) = P(Z < –0.3551) ≈ 0.361。因此检验功效为 0.639。
9. The Trade-off Between Errors | 两类错误间的权衡
There is an inherent trade-off: for a fixed sample size, decreasing α (making the test more conservative) will increase β, and vice versa. The only way to simultaneously reduce both error probabilities is to increase the sample size, which makes the test statistic more sensitive to departures from H₀. In exam questions, students may be asked to explain why a larger sample reduces the probability of a Type II error.
两者之间存在着内在的权衡：在样本量固定的情况下，减小 α（使检验更加保守）会增大 β，反之亦然。同时降低两类错误概率的唯一方法是增大样本容量，这会让检验统计量对偏离 H₀ 更加敏感。在考题中，学生可能会被要求解释为什么更大的样本能减少第二类错误的概率。
10. Calculating Type II Error for Two-Tailed Tests | 双尾检验中第二类错误的计算
c falls between the two critical values under the true alternative parameter. For example, with H₀: p = 0.5, H₁: p ≠ 0.5, α = 0.05, and n = 100, the critical region for the sample proportion p̂ under the normal approximation is p̂ ≤ 0.5 – 1.96√0.5×0.5/100 and p̂ ≥ 0.5 + 1.96√0.5×0.5/100. If the true p = 0.6, β = P(not rejecting | p = 0.6) = P(0.402 < p̂ < 0.598 | ge when the true value is in the direction of the tails but still close to the null value.
对于双尾检验，临界域被分割在两侧尾部。计算 β 时需要求出在真实的备择参数值下，检验统计量落在两个临界值之间的概率。例如，H₀: p = 0.5，H₁: p ≠ 0.5，α = 0.05，n = 100，基于正态近似，样本比例 p̂ 的临界域为 p̂ ≤ 0.5 – 1.96√0.5×0.5/100 和 p̂ ≥ 0.5 + 1.96√0.5×0.5/100。若真实 p = 0.6，利用实际分布 N(0.6, 0.6×0.4/100)，β = P(未拒绝 | p = 0.6) = P(0.402 < p̂ < 0.598 | p = 0.6)。这一计算表明，当真实值偏向某一尾部但仍然接近原假设值时，β 可能会很大。
11. Common Pitfalls and Exam Advice | 常见错误与考试建议
Edexcel examiners often test the precise definition of Type I and Type II errors in context. Students must avoid confusing the two. A common mistake is saying 'accept H₀' instead of 'do not reject H₀'. Also, when asked to define a Type II error, the answer must mention that H₀ is false. For calculations, always state the distribution of the test statistic under both H₀ and the given alternative, and be careful with continuity corrections when using the normal approximation to the binomial.
Edexcel 考官经常在给定情境下考查两类错误的精确定义。学生必须避免混淆两者。常见的错误是说“接受 H₀”而非“不拒绝 H₀”。此外，当被要求定义第二类错误时，答案必须指出 H₀ 是假的。在计算方面，务必同时写出检验统计量在 H₀ 和给定备择假设下的分布，并且在使用正态近似二项分布时注意连续性校正。
12. Conclusion and Key Takeaways | 总结与要点
Mastering Type I and Type II errors is essential for a deep understanding of hypothesis testing in Edexcel A-Level Mathematics. Remember: a Type I error is linked to α and occurs when you reject a true H₀; a Type II error depends on β and happens when you fail to reject a false H₀. The power of the test is 1 – β. Practice both binomial and normal examples, and always interpret errors in the context of the problem. By balancing α, β, and sample size, statisticians design experiments that minimise the risk of incorrect conclusions.
掌握第一类错误和第二类错误对于深入理解 Edexcel A-Level 数学中的假设检验至关重要。请记住：第一类错误与 α 相关，发生在拒绝真实 H₀ 时；第二类错误取决于 β，发生在未能拒绝错误 H₀ 时。检验功效为 1 – β。要多加练习二项分布和正态分布的示例，并在问题情境中解读错误含义。通过平衡 α、β 和样本量，统计学家能够设计出将错误结论风险降至最低的实验。
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