📚 In-Depth Analysis of Past Papers: Pre-U CAIE Statistics | Pre-U CAIE 统计：历年真题深度解析
Pre-U CAIE Statistics (9795) demands a blend of theoretical understanding and applied problem-solving. Past paper analysis reveals recurring themes, specific examiner expectations, and common pitfalls. This article dissects key topics using authentic-style questions, offering bilingual insights to sharpen your exam technique and deepen conceptual grasp.
Pre-U CAIE 统计（9795）要求理论理解与应用解题相结合。历年真题分析揭示了反复出现的主题、具体的考官预期以及常见陷阱。本文通过真题风格的问题剖析关键Topic，提供双语洞察，以提升你的考试技巧并加深概念掌握。
1. Exam Structure and Key Assessment Objectives | 考试结构与核心考核目标
Paper 1 (Probability and Statistical Inference, 3 h) assesses probability axioms, distributions, estimation, and hypothesis testing. Paper 2 (Applied and Computational Statistics, 3 h) focuses on modelling, data analysis, interpretation, and use of statistical tables or software output.
试卷一（概率与统计推断，3 小时）考查概率公理、分布、估计与假设检验。试卷二（应用与计算统计，3 小时）侧重建模、数据分析、解释以及统计表或软件输出的使用。
Examiners consistently test the ability to choose the correct model, justify assumptions, and communicate conclusions in context. Marks are often lost when candidates omit units, fail to define parameters, or present disorganised calculations.
考官始终考查选择正确模型、论证假设并结合背景给出结论的能力。考生常因遗漏单位、未定义参数或计算过程混乱而失分。
2. Probability, Tree Diagrams and Bayes’ Theorem | 概率、树状图与贝叶斯定理
Conditional probability P(A|B) = P(A ∩ B)/P(B) forms the backbone of many Paper 1 problems. Candidates are regularly asked to complete a tree diagram with probabilities, then find an intersection or a reverse conditional.
条件概率 P(A|B) = P(A ∩ B)/P(B) 是试卷一中许多问题的基石。考生经常被要求完善带有概率的树状图，然后求交集或反向条件。
A typical past-paper scenario: A factory produces components on two machines. Machine X produces 60% with a 5% defect rate; Machine Y produces the rest with a 3% defect rate. Find the probability a randomly chosen component is defective, and given it is defective, the probability it came from Machine Y. Bayes’ Theorem gives P(Y|D) = [P(D|Y)P(Y)] / [P(D|X)P(X) + P(D|Y)P(Y)] = (0.03×0.40)/(0.05×0.60+0.03×0.40).
一道典型的真题场景：一家工厂使用两台机器生产部件。机器 X 生产 60% 且缺陷率 5%；机器 Y 生产其余部分且缺陷率 3%。求随机选取的部件有缺陷的概率，以及已知部件有缺陷时它来自机器 Y 的概率。贝叶斯定理给出 P(Y|D) = [P(D|Y)P(Y)] / [P(D|X)P(X) + P(D|Y)P(Y)] = (0.03×0.40)/(0.05×0.60+0.03×0.40)。
Top answers clearly label events, draw a diagram, and present a fraction as the final answer before decimal approximation. Always check that conditional probabilities sum to 1 after the tree.
高分答案会清晰地标注事件，绘制图表，并在给出小数近似值前以分数呈现最终答案。务必检查树状图之后的条件概率之和是否为 1。
3. Discrete Random Variables: Binomial and Poisson | 离散随机变量：二项分布与泊松分布
Binomial distribution B(n, p) and Poisson distribution Po(λ) appear in both papers. Examiners test recognition of conditions, probability calculations using formula P(X=k) = ⁿCₖ pᵏ(1-p)ⁿ⁻ᵏ, and approximations such as Poisson to normal.
二项分布 B(n, p) 和泊松分布 Po(λ) 在两份试卷中都会出现。考官考查对条件的识别、使用公式 P(X=k) = ⁿCₖ pᵏ(1-p)ⁿ⁻ᵏ 计算概率，以及泊松近似正态等方法。
A frequent trap: using binomial when independence is violated. When a sample is taken without replacement from a finite population, the hypergeometric situation arises, but for large populations binomial may still be used as an approximation if the sample size is less than 10% of the population.
常见陷阱：在独立性被破坏时仍使用二项分布。从有限总体中无放回抽样时，应使用超几何分布，但如果样本量小于总体的 10%，对于大总体仍可用二项分布作为近似。
Poisson processes: the number of calls to a call centre per minute may be modelled as Po(2.5). Past questions ask for P(X > 3), which requires 1 − P(X ≤ 3). Accurate use of the cumulative Poisson table is essential, but also showing the working with addition of individual probabilities earns method marks.
泊松过程：每分钟打到呼叫中心的电话数量可建模为 Po(2.5)。真题会要求 P(X > 3)，需要 1 − P(X ≤ 3)。准确使用泊松累积分布表至关重要，但展示单个概率相加的推导过程也能获得方法分。
4. Continuous Random Variables and the Normal Distribution | 连续随机变量与正态分布
The normal distribution N(μ, σ²) is central. Past papers demand standardisation z = (x − μ)/σ, use of the standard normal table, and working backwards from a given probability to find μ or σ. For sample means, the standard error becomes σ/√n.
正态分布 N(μ, σ²) 处于核心地位。真题要求标准化 z = (x − μ)/σ、使用标准正态表，并从给定概率反推 μ 或 σ。对于样本均值，标准误变为 σ/√n。
A classic question: The weight of apples is normally distributed with mean 150 g and standard deviation 15 g. Find the proportion weighing more than 170 g. If five apples are selected at random, find the probability that their mean weight exceeds 170 g. The second part uses σ/√n = 15/√5, and the distribution of the sample mean remains normal.
经典问题：苹果重量呈正态分布，均值 150 g，标准差 15 g。求重量超过 170 g 的比例。若随机选取五个苹果，求其平均重量超过 170 g 的概率。第二部分需用 σ/√n = 15/√5，样本均值的分布仍是正态。
Candidates often forget to use the continuity correction factor when a discrete distribution (binomial/Poisson) is approximated by the normal. The correction 0.5 should be added or subtracted appropriately from the discrete value.
考生经常忘记在用正态分布近似离散分布（二项/泊松）时使用连续性修正因子。需在离散值上适当加减 0.5。
5. Sampling Distributions and the Central Limit Theorem | 抽样分布与中心极限定理
When n is large (n ≥ 30), the CLT states that the sample mean X̄ is approximately normally distributed with mean μ and variance σ²/n, regardless of the population distribution. This justifies many hypothesis tests and confidence intervals.
当 n 较大（n ≥ 30）时，中心极限定理指出样本均值 X̄ 近似服从均值为 μ、方差为 σ²/n 的正态分布，无论总体分布如何。这为许多假设检验和置信区间提供了依据。
Past papers often include a scenario where population is clearly non-normal, e.g., a skewed income distribution. Candidates must explicitly state that the CLT applies because the sample size is sufficiently large and then proceed with the normal distribution for the sample mean.
历年真题常包括总体明显非正态的场景，例如偏态的收入分布。考生必须明确陈述因为样本量足够大所以适用中心极限定理，然后对样本均值使用正态分布。
When population variance is unknown and n is small, the t-distribution with ν = n−1 degrees of freedom is required. Confusion between z and t is a common mark-losing error in confidence intervals and tests.
当总体方差未知且 n 较小时，需使用自由度为 ν = n−1 的 t 分布。在置信区间和检验中混淆 z 与 t 是常见的失分错误。
6. Hypothesis Testing for a Mean (Normal Population) | 正态总体均值的假设检验
Standard normal test (z-test) for mean with known variance: test statistic z = (x̄ − μ₀)/(σ/√n). Past papers ask for both critical region and p-value approaches. The null hypothesis H₀ is assumed true unless evidence suggests otherwise.
已知方差时均值的标准正态检验（z 检验）：检验统计量 z = (x̄ − μ₀)/(σ/√n)。真题要求掌握临界区域法和 p 值法。除非证据表明相反，否则认为原假设 H₀ 为真。
In a two-tailed test at 5% significance level, the critical values are ±1.96. If the calculated z falls beyond these, reject H₀. For a p-value approach, compare p with 0.05. Examiners expect a clear statement: “Reject H₀” or “Do not reject H₀”, together with a contextual conclusion.
在 5% 显著性水平的双侧检验中，临界值为 ±1.96。若计算的 z 超出此范围，则拒绝 H₀。对于 p 值法，将 p 与 0.05 比较。考官期待清晰的陈述：“拒绝 H₀”或“不拒绝 H₀”，并结合上下文的结论。
Type I error (rejecting true H₀) and Type II error (failing to reject false H₀) are explicitly tested. Candidates must be able to define these in the context of the question and calculate the probability of a Type II error when specific alternative values are given.
第一类错误（拒绝真实的 H₀）和第二类错误（未能拒绝错误的 H₀）会被明确考查。考生必须能在题目背景下定义它们，并在给出具体的备择值时计算第二类错误的概率。
7. Confidence Intervals for the Population Mean | 总体均值的置信区间
A 95% confidence interval for μ when σ is known: x̄ ± 1.96 × σ/√n. The interpretation is crucial: if we repeated the sampling many times, 95% of the constructed intervals would contain the true μ.
已知 σ 时 μ 的 95% 置信区间：x̄ ± 1.96 × σ/√n。其解释至关重要：若多次重复抽样，则构建的区间中有 95% 会包含真实的 μ。
Past questions often ask for the minimum sample size required to achieve a certain margin of error. Rearranging the formula, n ≥ (z × σ / margin)², and rounding up to the next integer. This type of question appears almost every year.
真题常要求达到指定边际误差所需的最小样本量。通过改写公式 n ≥ (z × σ / 误差)²，并向上取整到下一个整数。这类问题几乎每年都会出现。
Confidence intervals for proportions are also common: p̂ ± z√[p̂(1−p̂)/n], with the condition that np̂ ≥ 5 and n(1−p̂) ≥ 5. Candidates must check these conditions before using the normal approximation.
比例的置信区间也很常见：p̂ ± z√[p̂(1−p̂)/n]，条件为 np̂ ≥ 5 且 n(1−p̂) ≥ 5。考生在使用正态近似前必须检验这些条件。
8. Chi-Squared Tests: Goodness-of-Fit and Independence | 卡方检验：拟合优度与独立性
Chi-squared statistic: χ² = Σ[(O − E)² / E]. For goodness-of-fit, degrees of freedom ν = k − 1, where k is the number of categories; for independence in an r×c contingency table, ν = (r−1)(c−1).
卡方统计量：χ² = Σ[(O − E)² / E]。对于拟合优度，自由度 ν = k − 1，k 为类别数；对于 r×c 列联表的独立性检验，ν = (r−1)(c−1)。
Examiners require expected frequencies to be at least 5 for each cell, otherwise adjacent categories should be combined. A neatly presented table with observed, expected, (O−E), (O−E)², and (O−E)²/E earns full marks. The final decision compares the calculated χ² with the critical value from tables.
考官要求每个单元格的期望频数至少为 5，否则应合并相邻类别。整洁地呈现包含观察值、期望值、(O−E)、(O−E)² 和 (O−E)²/E 的表格可获满分。最终决策需将计算的 χ² 与查表临界值相比较。
Common mistake: using percentages or proportions instead of frequencies. Another is computing the wrong degrees of freedom, especially after combining categories. Always recount the number of categories after merging.
常见错误：使用百分比或比例而非频数。另一个错误是计算错误的自由度，尤其是在合并类别之后。合并后务必重新计算类别数量。
9. Linear Regression and Correlation | 线性回归与相关性
The least squares regression line y = a + bx is calculated using the summary statistics Σx, Σy, Σx², Σxy and n. Coefficients: b = [nΣxy − (Σx)(Σy)] / [nΣx² − (Σx)²], a = ȳ − b x̄. Interpretation of the slope b must be in context.
最小二乘回归线 y = a + bx 使用汇总统计量 Σx, Σy, Σx², Σxy 和 n 计算。系数：b = [nΣxy − (Σx)(Σy)] / [nΣx² − (Σx)²]，a = ȳ − b x̄。斜率的解释需结合具体背景。
Product-moment correlation coefficient r measures strength and direction of linear relationship. Hypothesis test for ρ using t = r√(n−2) / √(1−r²) with ν = n−2 is frequently examined. A high magnitude of r does not imply causation.
积矩相关系数 r 衡量线性关系的强度和方向。关于 ρ 的假设检验使用 t = r√(n−2) / √(1−r²)，自由度 ν = n−2，常被考查。r 的数值高并不意味着因果关系。
Predicting y for a given x requires caution: interpolation within the data range is acceptable, but extrapolation beyond the range is unreliable and must be stated as such. Residual analysis may be asked to assess model appropriateness.
对给定 x 预测 y 需谨慎：在数据范围内的内插是可接受的，但在范围外的外推不可靠且必须声明。可能会要求进行残差分析以评估模型适用性。
10. Common Pitfalls and Effective Exam Strategy | 常见失分点与高效应试策略
Many candidates lose time by overcomplicating simple probability problems. Drawing a quick tree or table often saves time and reduces errors. Always define variables and state assumptions before launching into calculations.
许多考生因将简单的概率问题过度复杂化而浪费时间。快速绘制树状图或表格通常能节省时间并减少错误。在开始计算前，务必定义变量并陈述假设条件。
Write formulas in symbols first, then substitute numbers. In hypothesis testing, always include a final concluding sentence in the context of the problem: “There is sufficient evidence, at the 5% level, to suggest that the mean weight has increased.”
先用符号写出公式，再代入数字。在假设检验中，务必在问题背景下写出最终结论句：“在 5% 显著性水平下，有充分证据表明平均重量增加了。”
Marks are awarded for clear working; even if the final answer is wrong, method marks can be earned. Time management: allocate about 1 minute per mark, and do not get stuck on one sub-question. The final questions often test synthesis — linking probability distributions with confidence intervals or tests. Practise mixed-topic past paper sections to build fluency.
步骤清晰可获得步骤分；即使最终答案错误，也能得到方法分。时间管理：大约每分标记 1 分钟，不要卡在一道小题上。最后几问通常测试综合能力——将概率分布与置信区间或检验联系起来。练习混合主题的历年真题部分以提升熟练度。
Published by TutorHao | Statistics 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