📚 In-depth Analysis of CIE Statistics Past Papers | CIE 统计历年真题深度解析
Mastering CIE A-Level Statistics requires not only understanding the theory but also applying it to the nuanced, context-rich questions that appear year after year. This article draws on real past paper patterns to demonstrate how to dissect typical problems, avoid recurrent mistakes, and write convincing, mark-winning solutions. Each section focuses on a core topic, presenting an adapted exam-style question and a step-by-step guide to a model answer.
要想在 CIE A-Level 统计中取得高分，不仅要理解理论，更要能应对每年真题中那些充满情境细节、考法多变的问题。本文依托历年真题的出题模式，通过改编的典型考题，展示如何拆解题目、规避常见失分点，并写出精准、拿下所有得分点的解答。每个小节聚焦一个核心考点，逐一深度解析。
1. The Poisson Distribution in Context | 泊松分布的情境化应用
One of the most frequent topics in CIE Statistics 2 is the Poisson distribution. Questions often embed it in a real-world scenario, such as the number of calls arriving at a switchboard or flaws in a length of fabric. A past paper might ask: ‘The number of emergencies per night in a hospital follows Po(2.5). Find the probability that in a randomly chosen week (7 nights) more than 20 emergencies occur.’ The key is to recognise that the sum of independent Poisson variables is again Poisson, allowing a single Poisson model with mean 7 × 2.5 = 17.5. Then either cumulative tables or a normal approximation with continuity correction must be used, depending on the marks available and instructions.
CIE 统计 2 最高频的考点之一是泊松分布。题目常把模型嵌入真实场景，比如某接线总机接到的呼叫次数，或一匹布料上的疵点个数。有一道真题可能这样问：“某医院每晚突发事件数服从 Po(2.5)。求随机选取的一周（7 晚）中突发事件超过 20 起的概率。”关键是要意识到独立泊松变量的和仍服从泊松分布，因此可以用均值为 7 × 2.5 = 17.5 的单一泊松模型处理。接着要根据分值及题目要求，选择使用累积分布表，或带有连续性校正的正态近似来计算。
Always write down the distribution you are using explicitly: ‘Let Y ~ Po(17.5) represent the weekly number of emergencies.’ Then state the required probability as P(Y > 20) = 1 – P(Y ≤ 20). If using a normal approximation, set μ = 17.5, σ² = 17.5, and apply P(Y > 20) ≈ P(Z > (20.5 – 17.5)/√17.5). The continuity correction is essential – many marks are lost by forgetting it. Conclude with a sentence that interprets the probability in context.
始终要把所引用的分布写清楚：“设 Y ~ Po(17.5) 表示一周的突发事件数。”然后将所求概率表示为 P(Y > 20) = 1 – P(Y ≤ 20)。若使用正态近似，令 μ = 17.5, σ² = 17.5，并应用 P(Y > 20) ≈ P(Z > (20.5 – 17.5)/√17.5)。连续性校正必不可少——许多考生因遗漏它而丢分。最后用一句陈述将概率放回语境中解释。
2. Approximating the Binomial: Poisson and Normal | 二项分布的逼近：泊松与正态
Past paper questions frequently test the conditions under which a binomial distribution B(n, p) can be approximated by a Poisson (np ≤ 5, n large) or by a normal distribution (np > 5, nq > 5). A typical problem provides a binomial probability and asks which approximation is suitable, followed by a calculation and comparison. For instance: ‘In a large sample of 200 items, the probability a single item is defective is 0.02. Using an appropriate approximation, find the probability that at most 3 items are defective.’ Here n = 200, p = 0.02, np = 4, so Poisson is ideal. The binomial P(X ≤ 3) can be approximated by Po(4) P(X ≤ 3) = 0.4335 from tables.
历年真题常考查二项分布 B(n, p) 在什么条件下可用泊松分布（np ≤ 5 且 n 很大）或正态分布（np > 5, nq > 5）来逼近。一道典型题目给出一个二项概率，问哪种近似合适，然后要求计算并比较。例如：“在 200 件大样本中，单个零件有缺陷的概率是 0.02，使用恰当的近似求至多有 3 件缺陷品的概率。”这里 n = 200, p = 0.02, np = 4，因此泊松分布是理想选择。原二项概率 P(X ≤ 3) 可由 Po(4) 的 P(X ≤ 3) = 0.4335 近似（查表）。
When the normal approximation is used, a continuity correction is again mandatory. For B(100, 0.35), np = 35, npq = 22.75, one must calculate P(34.5 < X < 35.5) or similar. Many CIE questions explicitly state 'using a suitable approximation' and expect you to justify your choice with a brief sentence like 'np = ... so the normal approximation is appropriate.' Don't skip the justification: it often carries a mark.
当使用正态近似时，同样必须应用连续性校正。对于 B(100, 0.35)，np = 35, npq = 22.75，计算时要写成 P(34.5 < X < 35.5) 等形式。很多 CIE 题目明确要求“用合适的近似”，并期望你用一句话说明选择理由，如“np = ... 所以正态近似合适”。切不可省略这一理由——这往往有分。
3. Continuous Random Variables: PDF and CDF | 连续随机变量：概率密度与累积分布
Questions on continuous distributions frequently require you to find the constant k in a probability density function (PDF), derive the cumulative distribution function (CDF), and compute probabilities such as P(X > a) or the median. For example: ‘The continuous random variable X has PDF f(x) = k(4 – x²) for –2 ≤ x ≤ 2, and 0 otherwise. (a) Find k. (b) Find the CDF F(x). (c) Determine the lower quartile.’ To solve (a), integrate f(x) over the domain and set equal to 1. The integral of (4 – x²) from –2 to 2 is 32/3, so k = 3/32.
连续分布的相关题目经常要求你求出概率密度函数（PDF）中的常数 k，导出累积分布函数（CDF），并计算诸如 P(X > a) 或中位数的概率。例如：“连续随机变量 X 的 PDF 为 f(x) = k(4 – x²)，–2 ≤ x ≤ 2，其余为 0。(a) 求 k。(b) 求 CDF F(x)。(c) 确定下四分位数。”要解 (a)，在定义域上积分 f(x) 并令其等于 1。∫(4 – x²) dx 从 –2 到 2 为 32/3，故 k = 3/32。
A common pitfall is forgetting to specify the range for the CDF: F(x) must be defined piecewise, with F(x) = 0 for x < –2, F(x) = the integral for –2 ≤ x ≤ 2, and F(x) = 1 for x > 2. When finding quartiles, set F(q) = 0.25 and solve within the appropriate domain. Always verify that your solution lies in the valid interval; if not, reject it. CIE mark schemes rigorously check these piecewise statements.
一个常见错误是忘记为 CDF 写明分段范围：F(x) 必须是分段定义的，即 x < –2 时 F(x) = 0，–2 ≤ x ≤ 2 时为积分结果，x > 2 时 F(x) = 1。求四分位数时，令 F(q) = 0.25 并在相应定义域内求解。始终要验证解是否落在有效区间内；若不在，则舍去。CIE 的评分标准严格核查这些分段陈述。
4. Expectation and Variance of Continuous Variables | 连续变量的期望与方差
Once the PDF is known, expectation and variance are computed via integration. For the same PDF f(x) = (3/32)(4 – x²), E(X) = ∫ x f(x) dx over the domain. Since the function is even, the expected value is 0 by symmetry. Variance E(X²) is ∫ x² f(x) dx, which evaluates to 0.8 after integration. In CIE past papers, you may be asked to find E(X²) directly or to verify that Var(X) = 0.8. The process requires careful integration and algebraic simplification; marks are awarded for correct limits and integral setup, not just the final answer.
一旦得到 PDF，期望和方差就通过积分来计算。对于相同的 PDF f(x) = (3/32)(4 – x²)，E(X) = ∫ x f(x) dx 在定义域上积分。由于函数为偶函数，由对称性可得期望为 0。方差 E(X²) = ∫ x² f(x) dx，积分后结果为 0.8。在 CIE 真题中，可能直接要求你求出 E(X²)，或验证 Var(X) = 0.8。过程要求细心积分与代数化简；得分点不仅在于最终答案，还包括正确的积分限与积分式的建立。
Always show the step where you expand the integrand: x² (4 – x²) = 4x² – x⁴. Then integrate term by term: [ (4/3)x³ – (1/5)x⁵ ]. Evaluate between –2 and 2, taking advantage of symmetry to double the integral from 0 to 2. Write the working clearly, as method marks heavily outweigh final answer marks when only a minor slip occurs.
务必展示展开被积函数的步骤：x² (4 – x²) = 4x² – x⁴。然后逐项积分：[ (4/3)x³ – (1/5)x⁵ ]。在 –2 到 2 之间求值，并利用对称性将 0 到 2 的积分值乘以 2。清晰写出步骤，因为当仅出现微小失误时，方法分的比重远大于最终答案分。
5. Linear Combinations of Independent Random Variables | 独立随机变量的线性组合
Another staple of S2 is the linear combination of independent normal variables. A question might say: ‘The masses of apples are normally distributed with mean 150 g and variance 25 g². The masses of oranges are normally distributed with mean 120 g and variance 16 g². A bag contains 3 apples and 2 oranges. Find the probability that the total mass exceeds 660 g.’ The total mass T = 3A + 2O, where A ~ N(150,25) and O ~ N(120,16). Assuming independence, T ~ N(3×150 + 2×120, 9×25 + 4×16) = N(690, 289). Then P(T > 660) is standardised and evaluated.
S2 的另一核心考点是独立正态变量的线性组合。题目可能是：“苹果的重量服从正态分布，均值为 150 g，方差为 25 g²；橙子的重量服从正态分布 N(120, 16)。一个袋子装有 3 个苹果和 2 个橙子，求总重量超过 660 g 的概率。”总重量 T = 3A + 2O，其中 A ~ N(150,25)，O ~ N(120,16)。假设独立，则 T ~ N(3×150 + 2×120, 9×25 + 4×16) = N(690, 289)。于是 P(T > 660) 经标准化后求值。
A common error is to forget that the variance multiplies by the square of the coefficient: Var(3A) = 9Var(A). Another subtlety appears when variables are not independent; exam questions will either explicitly state independence or provide the necessary covariance. When given the sum X + Y with known Var(X + Y) = Var(X) + Var(Y) + 2Cov(X,Y), you must incorporate this. Always check the wording.
常见错误是忘记方差要乘以系数的平方：Var(3A) = 9Var(A)。另一个微妙的点出现在变量不独立时；考题会明确说明独立性，或给出必要的协方差。若给出 X + Y 且已知 Var(X + Y) = Var(X) + Var(Y) + 2Cov(X,Y)，必须将其纳入。务必留意题目表达。
6. Sampling Distributions and the Central Limit Theorem | 抽样分布与中心极限定理
The Central Limit Theorem (CLT) is frequently examined, especially when the sample size n is large and the population distribution is non-normal. A typical past paper question: ‘The breaking strength of a cable has mean 500 N and standard deviation 30 N. A random sample of 64 cables is taken. Find the probability that the sample mean strength is less than 495 N.’ By the CLT, the sample mean X̄ ~ N(500, 30²/64) approximately, even if the population is not normal. Then compute P(X̄ < 495) = P(Z < (495 – 500)/(30/√64)) = P(Z < –1.333) ≈ 0.0912. The CLT justification is a required step: 'since n = 64 is large, the distribution of the sample mean is approximately normal.'
中心极限定理（CLT）是高频考点，尤其是当样本容量 n 很大且总体非正态时。一道典型真题：“某缆绳的断裂强度均值为 500 N，标准差为 30 N。随机抽取 64 根缆绳组成样本，求样本均值低于 495 N 的概率。”根据中心极限定理，X̄ ~ N(500, 30²/64) 近似，即使总体非正态。然后计算 P(X̄ < 495) = P(Z < (495 – 500)/(30/√64)) = P(Z < –1.333) ≈ 0.0912。CLT 的说明是必备步骤：“由于 n = 64 足够大，样本均值的分布近似正态。”
Sometimes the question provides a proportion: ‘A biased coin is tossed 100 times. Estimate the probability that the proportion of heads exceeds 0.6.’ The sample proportion p̂ is approximately N(p, p(1–p)/n) by CLT. Always check if a continuity correction is needed—typically for proportions, it may be ignored when n is large and the question asks for an estimate, but it’s safer to consider. CIE often awards marks for stating the approximate distribution correctly.
有时题目会采用比例形式：“抛一枚偏硬币 100 次，估计正面比例超过 0.6 的概率。”根据 CLT，样本比例 p̂ 近似服从 N(p, p(1–p)/n)。务必判断是否需要连续性校正——当 n 很大且题目要求估计时，常可忽略，但保守起见仍应考虑。CIE 通常会对正确写出近似分布给出分数。
7. Confidence Intervals for the Mean | 均值的置信区间
Constructing a confidence interval for a population mean is a standard past paper task. A question may state: ‘A sample of 25 bolts has a mean diameter of 10.2 mm. The population standard deviation is known to be 0.4 mm. Calculate a 95% confidence interval for the true mean diameter.’ The formula is x̄ ± z × σ/√n, where z = 1.96 for 95% confidence. This gives 10.2 ± 1.96 × 0.4/5 = (10.043, 10.357). Interpretation must be in context: ‘We are 95% confident that the true mean diameter lies between 10.04 mm and 10.36 mm.’
构建总体均值的置信区间是真题中的常规任务。题目会给出：“一个 25 颗螺栓的样本平均直径为 10.2 mm，已知总体标准差为 0.4 mm。计算总体真实平均直径的 95% 置信区间。”公式为 x̄ ± z × σ/√n，其中 95% 置信水平下 z = 1.96。得 10.2 ± 1.96 × 0.4/5 = (10.043, 10.357)。解释必须夹带情境：“我们有 95% 的把握认为真实平均直径介于 10.04 mm 到 10.36 mm 之间。”
When σ is unknown and the sample size is small, the t-distribution is required. CIE past papers frequently test the t-interval with n ≤ 30. You must find the critical t-value with n–1 degrees of freedom from the table. For a 99% interval with n = 10, ν = 9, t = 3.250. The interval is x̄ ± t × s/√n. Failure to switch from z to t in small samples is a classic error.
当 σ 未知且样本较小时，必须使用 t 分布。CIE 真题经常考查 n ≤ 30 时的 t 区间。须根据自由度 n–1 从表中查出临界 t 值。例如 n = 10 的 99% 置信区间，ν = 9，t = 3.250。区间公式为 x̄ ± t × s/√n。小样本时忘记从 z 转向 t 是经典扣分点。
8. Hypothesis Testing for the Mean (Known Variance) | 已知方差的均值假设检验
A typical exam question: ‘A manufacturer claims the mean lifetime of a battery is 50 hours. A sample of 36 batteries yields x̄ = 48.5 h. Assume σ = 5 h. Test the manufacturer’s claim at the 5% significance level.’ Set up H₀: μ = 50, H₁: μ < 50. The test statistic is Z = (48.5 – 50)/(5/√36) = –1.8. The critical value for a one-tailed test at α = 0.05 is –1.645. Since –1.8 < –1.645, we reject H₀ and conclude there is sufficient evidence that the mean lifetime is below 50 hours. Always state 'reject H₀' or 'do not reject H₀', and tie back to the context.
一道典型真题：“某厂商声称其电池平均寿命为 50 小时。抽取 36 块电池得 x̄ = 48.5 h，已知 σ = 5 h。在 5% 显著性水平下检验厂商的说法。”设 H₀: μ = 50，H₁: μ < 50。检验统计量 Z = (48.5 – 50)/(5/√36) = –1.8。单侧检验 α = 0.05 下的临界值为 –1.645。由于 –1.8 < –1.645，拒绝 H₀，得出结论：有充分证据表明平均寿命低于 50 小时。务必明确写出“拒绝 H₀”或“不拒绝 H₀”，并联系题设背景。
When the alternative is two-sided, the critical region is split equally. For α = 0.05, the critical values are ±1.96. Compare the absolute test statistic to 1.96. Don’t forget to define the rejection rule clearly: ‘Reject H₀ if |Z| > 1.96.’ Some questions ask for the p-value instead: compute the tail probability and compare to α. The novelty in CIE is that they often require a conclusion written in non-statistical language for the last mark.
当备择假设为双侧时，拒绝域对半划分。α = 0.05 时，临界值为 ±1.96。将检验统计量的绝对值与 1.96 比较。务必明确定义拒绝法则：“若 |Z| > 1.96，则拒绝 H₀。”部分题目要求提供 p 值：计算尾部概率并与 α 比较。CIE 的特点在于，常需要用一个非统计语言的结论来拿下最后一分。
9. Type I and Type II Errors and Power | 两类错误与检验功效
CIE S2 past papers increasingly feature questions on errors and power. A typical scenario: ‘A test with H₀: μ = 100, H₁: μ > 100, σ = 10, n = 25, significance level 5%. Find the probability of a Type II error when the true mean is 102.’ First find the critical region in terms of x̄: reject if x̄ > 100 + 1.645×10/5 = 103.29. Then β = P(do not reject H₀ | μ = 102) = P(x̄ ≤ 103.29 | μ = 102). Standardise under the true distribution: Z = (103.29 – 102)/2 = 0.645, so β ≈ 0.740. Therefore power = 1 – β ≈ 0.260. These problems require careful structuring of two different normal curves.
CIE S2 真题越来越频繁地考查错误类型与功效。典型情景：“检验 H₀: μ = 100, H₁: μ > 100, σ = 10, n = 25, 显著性水平 5%。当真实均值为 102 时，求第 II 类错误的概率。”首先求出以 x̄ 表述的拒绝域：若 x̄ > 100 + 1.645×10/5 = 103.29，则拒绝 H₀。那么 β = P(不拒绝 H₀ | μ = 102) = P(x̄ ≤ 103.29 | μ = 102)。在真实分布下标准化：Z = (103.29 – 102)/2 = 0.645，故 β ≈ 0.740。因此功效 = 1 – β ≈ 0.260。此类问题需要细心构建两条不同的正态曲线。
Dividing the solution into clear sections helps: (1) state the critical value, (2) express the acceptance region, (3) compute probability for the alternative distribution. A diagram is often useful, but in text, label the regions explicitly. Marks are given for correct formulas and standardisation; an arithmetic slip should still earn method marks if the logic is clear.
将解答划分为清晰的部分会很有帮助：（1）陈述临界值；（2）写出接受域表达式；（3）在备择分布下计算概率。图示常有助益，但在文本解答中必须明确标出各区域。正确的公式与标准化过程会得分；即使出现计算错误，只要逻辑清晰，仍可获得方法分。
10. Chi-Squared Tests: Goodness of Fit and Contingency Tables | 卡方检验：拟合优度与独立性
Chi-squared problems appear almost every year. For goodness of fit: ‘Data on the number of absent days per employee: 0, 1, 2, and ≥3. Observed frequencies are 25, 12, 8, 5. Test at 5% if the data follow a Poisson distribution with mean 0.8.’ First find expected frequencies by computing Poisson probabilities for 0,1,2, and P(X ≥ 3) = 1 – P(X ≤ 2), then multiply by total N=50. Combine categories if any expected frequency < 5. Calculate χ² = Σ(O – E)²/E. Compare with critical χ² with (number of categories – 1 – estimated parameters) degrees of freedom. Here estimated parameter λ̂ = 0.8 (given), so ν = 4 – 1 – 1 = 2? Wait, categories: 0,1,2,≥3 = 4 categories. Estimated parameter is 0? No, given mean, so not estimated from data? If given directly, the degrees of freedom are categories – 1 = 3. Only when mean is estimated from data does one subtract an extra degree. Read carefully.
卡方检验几乎每年必考。拟合优度题：“每名员工的缺勤天数数据：0, 1, 2, 和 ≥3。观察频数为 25, 12, 8, 5。在 5% 显著性水平下检验数据是否服从均值为 0.8 的泊松分布。”首先通过计算泊松分布中 0, 1, 2 的概率，以及 P(X ≥ 3) = 1 – P(X ≤ 2)，再乘以总人数 N=50 得到期望频数。若某期望频数 < 5 则需合并类别。计算 χ² = Σ(O – E)²/E。与临界 χ² 比较，自由度为（类别数 – 1 – 估计参数个数）。此处参数 λ̂ = 0.8 是直接给出的，并非由数据估计，故自由度 = 4 – 1 = 3。如果均值是由样本数据估计的，则需再减 1，务必仔细审题。
For contingency tables, the null hypothesis is that two variables are independent. Expected frequencies are (row total × column total) / grand total. Degrees of freedom = (rows – 1)×(columns – 1). Always state hypotheses formally. After obtaining χ², state the conclusion in context: ‘There is evidence of an association between …’ or ‘No significant association …’. CIE often requires interpretation of residuals to indicate which cells contribute most to χ².
对于列联表，原假设为两变量独立。期望频数为（行合计 × 列合计）/ 总计。自由度 =（行数 – 1）×（列数 – 1）。务必正式写出假设。得到 χ² 后，在情境中给出结论：“有证据表明…之间存在关联”或“无显著关联…”。CIE 常要求通过分析残差来说明哪些单元格对 χ² 贡献最大。
11. Past Paper Strategy: Interpreting Context and “Write in Context” | 真题答题策略：情境化结论
One of the most common marks lost in CIE S2 is the final ‘interpretation in context’ mark. After computing a confidence interval or performing a hypothesis test, candidates often state ‘reject H₀’ and stop. The mark scheme demands a sentence that translates the statistical decision back to the original problem. For instance, after testing whether a new drug reduces recovery time, say: ‘There is sufficient evidence to conclude that the new drug reduces average recovery time.’ Not just ‘reject H₀’. Practise extracting the keywords from the question stem and weaving them into your conclusion.
CIE S2 中最容易丢掉的分数之一就是最后的“情境化解释”分。算出置信区间或完成假设检验后，考生往往只写出“拒绝 H₀”就停笔。但评分标准要求用一句话将统计决策转译为原问题。例如，在检验一种新药是否缩短恢复时间后，应写：“有充分证据表明新药降低了平均恢复时间。”而不只是“拒绝 H₀”。要练习从题干中提取关键词，并将其编织进你的结论中。
Another strategic point: show all working, even if the final answer is wrong. Use labelled steps: H₀, H₁, test statistic, critical value, comparison, decision, contextual conclusion. Clear structure not only earns method marks but also helps you avoid skipping a step. For calculation-heavy questions
Published by TutorHao | Year 13 统计 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