High-Frequency Topics and Common Mistakes in Pre-U OCR Statistics | Pre-U OCR 统计：高频考点与易错题分析

📚 High-Frequency Topics and Common Mistakes in Pre-U OCR Statistics | Pre-U OCR 统计：高频考点与易错题分析

The Pre-U OCR Statistics paper is designed to probe conceptual depth rather than mere mechanical computation. Students often perform well on routine probability calculations yet stumble on the interpretation of confidence intervals, the selection of appropriate hypothesis tests, and the precise conditions required for asymptotic approximations. This article surveys the topics that appear with striking regularity and highlights the subtle errors that cost marks year after year.

Pre-U OCR 统计考试旨在考查对概念的深层理解，而非简单的机械运算。考生往往能熟练完成常规的概率计算，却在置信区间的阐释、假设检验方法的选择以及渐近近似条件的验证上频频失分。本文梳理了反复出现的高频考点，并重点剖析那些年年都会让考生付出代价的细微错误。

1. Probability Distributions and Parameter Pitfalls | 概率分布与参数陷阱

A foundational block of the Pre-U syllabus is fluency with discrete and continuous random variables. Examiners love to embed a parameter (such as k) within a probability mass function and ask candidates to solve for it using the fact that total probability must sum to 1. A common mistake is to set up the equation correctly but misread the domain of the variable, thus missing terms or double-counting.

离散与连续随机变量是 Pre-U 大纲的基石。考官喜欢在概率质量函数中嵌入参数（如 k），要求考生利用总概率为 1 的性质求解。常见错误是方程列对，却误读了变量的取值范围，导致漏项或重复计数。

Another high-risk area is the difference between the probability density function (pdf) and cumulative distribution function (cdf) for continuous distributions. When computing P(a < X < b), students must integrate the pdf, not evaluate the pdf at a and b. Misplacing the density value for a probability is a classic slip.

另一个高风险区域是连续分布中概率密度函数（pdf）和累积分布函数（cdf）的区别。计算 P(a < X < b) 时，必须对 pdf 积分，而不是在 a 与 b 处取密度值。将密度值误当作概率是经典的疏忽。

∫ₐᵇ f(x) dx ≠ f(a) or f(b)

2. Binomial Distribution and Poisson Approximation | 二项分布与泊松近似

The binomial setting X ~ B(n, p) is tested relentlessly. Candidates often lose marks by failing to distinguish between P(X = x) and P(X ≤ x). When the problem asks for ‘more than 10 successes’, they write 1 – P(X ≤ 10) which is correct, but they occasionally use P(X < 10) incorrectly, forgetting that P(X > 10) = 1 – P(X ≤ 10). Another recurrent error is using the normal approximation without applying a continuity correction, or using the Poisson approximation when n is large but p is close to 0, without verifying that np < 10 and n is large enough.

二项分布 X ~ B(n, p) 是必考内容。考生经常因混淆 P(X = x) 与 P(X ≤ x) 而失分。当题目要求“超过 10 次成功”时，他们写 1 – P(X ≤ 10) 是正确的，但有时会错误地使用 P(X < 10)，忘记了 P(X > 10) = 1 – P(X ≤ 10)。另一个屡犯的错误是在未使用连续性校正的情况下直接进行正态近似，或者在 n 大但 p 接近 0 时使用泊松近似，却没有验证 np < 10 且 n 充分大。

The exact condition for the Poisson approximation to work well is that n is large, p is small, and np remains moderate. Many candidates simply state ‘n is large, p is small’ without checking the numerical bound, which leaves their reasoning incomplete.

泊松近似适用性的确切条件是 n 大、p 小且 np 处于中等水平。许多考生仅仅陈述“n 大 p 小”，却没有检查数值边界，这使他们的推理不够完整。

3. Normal Distribution and the Standardisation Trap | 正态分布与标准化陷阱

The normal distribution X ~ N(μ, σ²) is almost guaranteed to appear, but errors cluster around the distinction between variance (σ²) and standard deviation (σ). When standardising, the formula is Z = (X – μ) / σ, not dividing by σ². In questions where the variance is given, students often forget to take the square root, leading to an absurd Z-value. Conversely, when asked to find an unknown μ or σ from a given probability, they sometimes set up the Z-equation incorrectly: P(X > a) = 0.1 becomes (a – μ)/σ = ? instead of (a – μ)/σ = z₀.₁ with the appropriate sign.

正态分布 X ~ N(μ, σ²) 几乎必考，但错误往往集中在方差（σ²）与标准差（σ）的区别上。标准化公式为 Z = (X – μ) / σ，而非除以 σ²。当题目给出方差时，考生常忘记开平方根，得到荒诞的 Z 值。反之，当需要从给定概率求未知的 μ 或 σ 时，他们有时会错误地建立 Z 方程：P(X > a) = 0.1 被写成 (a – μ)/σ = ?，而不是带正确符号的 (a – μ)/σ = z₀.₁。

Another trap is the use of the standard normal table. If a table gives P(Z < z) and the question asks for the upper tail, many candidates forget to subtract from 1. They also confuse z₀.₀₅ = 1.645 for a one-tailed test with z₀.₀₂₅ = 1.96 for a two-tailed test at the 5% significance level.

另一个陷阱是标准正态表的使用。如果表格给出的是 P(Z < z)，而题目要求上尾概率，许多考生忘记用 1 去减。他们还混淆了单尾检验显著性水平 5% 的临界值 z₀.₀₅ = 1.645 与双尾检验 5% 的临界值 z₀.₀₂₅ = 1.96。

4. Sampling Distributions and the Central Limit Theorem | 抽样分布与中心极限定理

One of the most misinterpreted results is the Central Limit Theorem (CLT). Students often recite that ‘the sample mean is normally distributed’, but they forget the crucial condition: for large enough sample size, regardless of the population distribution. A typical mistake is applying the CLT to a single observation rather than to the mean of a sample. The distribution of the sample mean X̄ has mean μ and variance σ²/n, not σ². Many candidates correctly write down the variance but then use σ by mistake in the standard error.

中心极限定理（CLT）是最容易被误解的结论之一。学生们常常背诵“样本均值服从正态分布”，却忘了关键条件：在样本量足够大时，无论总体分布如何。典型的错误是将 CLT 应用于单个观测值，而非样本均值。样本均值 X̄ 的分布具有均值 μ 和方差 σ²/n，而不是 σ²。许多考生能正确写下方差，却在标准误中误用了 σ。

When the population variance is unknown and estimated by s², the correct distribution for the sample mean is t-distribution with n – 1 degrees of freedom, yet candidates frequently jump straight to the normal distribution. This is disastrous for small samples.

当总体方差未知且用 s² 估计时，样本均值的正确分布是自由度为 n – 1 的 t 分布，但考生常常直接跃入正态分布。这对于小样本而言是灾难性的。

5. Point Estimation: Maximum Likelihood Estimation | 点估计：最大似然估计

Maximum Likelihood Estimation (MLE) is a perennial Pre-U favourite. The log-likelihood function must be differentiated with respect to the parameter(s). Examiners report a cascade of errors: forgetting to take the log before differentiating, misapplying the product rule, or omitting the domain of the parameter (e.g., θ > 0). Another classic mistake is to find a stationary point but fail to verify it is a maximum, either by the second derivative test or by inspecting the endpoints.

最大似然估计（MLE）是 Pre-U 的常年热门。对数似然函数必须对参数求导。考官报告了一系列错误：忘记先取对数再求导、错误使用乘积法则，或遗漏参数的定义域（如 θ > 0）。另一个经典错误是求出驻点却没有验证它是最大值，或未通过二阶导数检验，或未检查端点。

In discrete cases, candidates sometimes write the likelihood as a product without indicating that it is valid only for the observed data set. For a sum of independent Poisson observations, the MLE of λ is simply the sample mean, but students who derive it often forget to use the joint probability mass function correctly, missing the factorial terms that vanish in differentiation.

在离散情况下，考生有时将似然函数写成乘积形式，却不标明它仅对观测数据集有效。对于独立泊松观测之和，λ 的 MLE 就是样本均值，但推导的学生常常忘记正确使用联合概率质量函数，遗漏了求导时会消失的阶乘项。

6. Confidence Intervals: Construction and Misinterpretation | 置信区间的构建与误解

The step-by-step construction of a confidence interval for a mean or proportion is well rehearsed, yet the interpretation is where marks bleed. A 95% confidence interval does not mean ‘there is a 95% probability that the true parameter lies in this particular interval’. The correct interpretation is: ‘If we were to take many independent random samples and construct a 95% confidence interval from each, then approximately 95% of those intervals would contain the true parameter value.’ The probability statement refers to the method, not to the specific interval once calculated.

均值或比例置信区间的逐步构建已经演练得很熟练，但阐释才是失分点。95% 置信区间并不意味着“真实参数有 95% 的概率落在这个特定区间内”。正确的解释是：“如果我们抽取许多独立随机样本，并从每个样本构造一个 95% 置信区间，那么大约 95% 的区间会包含真实参数值。” 概率陈述针对的是方法，而非计算出的特定区间。

A second common pitfall is choosing the wrong multiplier. For a mean with known population σ, use z*; with unknown σ, use t* with n – 1 degrees of freedom. For a proportion, the standard error is √[p̂(1 – p̂)/n] and the multiplier is always z* (based on the normal approximation, provided the success-failure condition holds). Mixing these up leads to invalid intervals.

第二个常见陷阱是选错乘数。对于均值，若总体 σ 已知，用 z*；若 σ 未知，用自由度为 n – 1 的 t*。对于比例，标准误为 √[p̂(1 – p̂)/n]，乘数始终为 z*（基于正态近似，前提是满足成功-失败条件）。混淆这些会导致无效的区间。

7. Hypothesis Testing: Pitfalls and Test Selection | 假设检验：常见错误与方法选择

The greatest source of confusion is choosing between a one-tailed and a two-tailed test. A two-tailed test is used when the alternative hypothesis is simply ‘not equal to’. If the research question predicts a direction (‘greater than’, ‘less than’), a one-tailed test is appropriate, but students often default to two-tailed, reducing the test’s power and sometimes failing to reject a false null. Conversely, using a one-tailed test when no direction is specified is equally invalid.

最大的混淆来源是单尾与双尾检验的选择。当备择假设仅仅是“不等于”时，使用双尾检验。如果研究问题预测了方向（“大于”、“小于”），则适用单尾检验，但学生常常默认双尾，这降低了检验的功效，有时导致未能拒绝错误的零假设。相反，在未指定方向时使用单尾检验同样无效。

The table below summarises common test scenarios and the corresponding conditions. Misapplication of parametric tests when the underlying distribution is severely non-normal or when the sample size is tiny is a frequent error.

下表总结了常见的检验情形及相应条件。当基础分布严重非正态或样本量极小时，错误地应用参数检验是一个频发失误。

Test	When to Use	Key Condition
One-sample z-test for mean	Population variance known or large n	Normality or CLT holds
One-sample t-test	Population variance unknown, small n	Underlying population ~ Normal
Two-sample t-test (unpaired)	Compare means of two independent groups	Both groups normal, equal variance assumed or checked
Paired t-test	Two measurements on the same subject	Differences ~ Normal
z-test for proportion	Binary outcome, large n	np₀ ≥ 10 and n(1–p₀) ≥ 10
Chi-squared goodness-of-fit	Categorical data, one sample	Expected frequencies ≥ 5

8. Chi-Squared Tests: Goodness of Fit and Independence | 卡方检验：拟合优度与独立性

Chi-squared tests appear in two guises: (i) goodness of fit, comparing observed frequencies to a theoretical distribution, and (ii) test for association/independence in a contingency table. A high-frequency error is using a goodness-of-fit test on raw data rather than summarised frequency data, or forgetting to combine categories when expected counts fall below 5. Without the merging step, the test statistic may not follow the χ² distribution adequately, invalidating the p-value.

卡方检验以两种形式出现：(i) 拟合优度检验，将观测频数与理论分布进行比较；(ii) 列联表中的独立性/关联性检验。一个高频错误是在原始数据上使用拟合优度检验，而非摘要频数数据，或者在期望计数低于 5 时忘记合并类别。若缺少合并步骤，检验统计量可能无法充分服从 χ² 分布，使 p 值失效。

In the test for independence, candidates often confuse the degrees of freedom. For an r × c contingency table, the df is (r – 1)(c – 1). A slip arises when they incorrectly use rc – 1 or (r – 1)(c – 1) – 1. Moreover, the calculation of expected frequencies: Eᵢⱼ = (row total × column total) / grand total is straightforward, yet simple arithmetical mistakes leave many otherwise strong answers marred.

在独立性检验中，考生常混淆自由度。对于一个 r × c 列联表，自由度为 (r – 1)(c – 1)。他们有时误用 rc – 1 或 (r – 1)(c – 1) – 1 而屡屡出错。此外，期望频数的计算：Eᵢⱼ = (行合计 × 列合计) / 总计 并不难，但简单的算术错误往往让本可出色的答案留下瑕疵。

9. Regression and Correlation Analysis | 回归与相关分析

Pearson’s product-moment correlation coefficient r measures the strength of linear association. A serious misunderstanding is to infer causation from a high |r|. Pre-U questions frequently ask for a comment, in context, and a bland statement such as ‘there is a strong positive correlation’ without a contextual interpretation will not earn full marks. Candidates must discuss what the correlation means for the variables in the given scenario and acknowledge lurking variables.

皮尔逊积矩相关系数 r 衡量线性关联的强度。一个严重的误解是从高 |r| 推断因果关系。Pre-U 题目经常要求结合上下文进行评论，仅仅平淡地陈述“存在强正相关”而不作情境解读，无法获得满分。考生必须讨论相关性对所给场景中变量的意义，并承认潜在变量的存在。

In linear regression y = a + bx, the least squares estimates are derived from the data. A common error is to mix up the slope and intercept. Another is to apply the regression equation for extrapolation far beyond the observed range of x without caution. Exam questions may ask for a prediction, expecting a comment that the prediction may be unreliable because it lies outside the data range.

在线性回归 y = a + bx 中，最小二乘估计来自数据。常见错误是混淆斜率和截距。另一个错误是在远离观测 x 范围的区域外推回归方程，而不加警示。考题可能要求作出预测，并期望附上评论：该预测可能不可靠，因为它超出了数据范围。

Also, when calculating the coefficient of determination R², it represents the proportion of variation in y explained by x. Students often misinterpret it as a measure of the steepness of the slope or the goodness of the fit in absolute residuals; it is a relative measure of explained variance.

此外，在计算决定系数 R² 时，它表示 y 的变异中由 x 解释的比例。考生常常将其误解为斜率陡峭程度的度量，或是绝对残差的拟合优度；它实际是解释方差的相对度量。

10. Bayes’ Theorem and Conditional Probability | 贝叶斯定理与条件概率

Bayes’ theorem questions often come with a twist: a disease testing scenario, or a factory production line. The fundamental error is confusing P(A|B) with P(B|A). The formula P(A|B) = [P(B|A) P(A)] / P(B) must be applied with precision. Students sometimes calculate P(B) incorrectly, forgetting that P(B) = P(B|A)P(A) + P(B|A’)P(A’). They might plug in the raw numbers without weighting by the prior probabilities, leading to a grossly inflated posterior probability.

A related pitfall is the identification of mutually exclusive and exhaustive events when setting up the law of total probability. When the sample space is partitioned into more than two events, candidates may omit a branch, or mislabel the conditional probabilities in a tree diagram.

相关的陷阱是在建立全概率公式时识别互斥且完备的事件。当样本空间被划分为两个以上的事件时，考生可能遗漏一个分支，或在树图中误标条件概率。

P(A|B) = [P(A ∩ B)] / P(B) = [P(B|A) P(A)] / [P(B|A)P(A) + P(B|A’)P(A’)]

11. Continuous Distributions: Exponential and Uniform | 连续分布：指数与均匀分布

The exponential distribution X ~ Exp(λ) with pdf f(x) = λe⁻ˣˣ, x ≥ 0 models waiting times. The memoryless property P(X > s + t | X > s) = P(X > t) is a favourite conceptual question. Examinees often miss that the rate parameter λ is the reciprocal of the mean; they might set up the mean as λ instead of 1/λ. Also, when finding the median, they solve F(m) = 0.5 but misapply the cdf F(x) = 1 – e⁻ˣˣ.

指数分布 X ~ Exp(λ) 的概率密度函数为 f(x) = λe⁻ˣˣ, x ≥ 0，用于建模等待时间。其无记忆性 P(X > s + t | X > s) = P(X > t) 是常见的概念题。考生常忽略速率参数 λ 是均值的倒数；他们可能将均值设为 λ 而非 1/λ。此外，在求中位数时，他们解方程 F(m) = 0.5 却误用累积分布函数 F(x) = 1 – e⁻ˣˣ。

For the continuous uniform distribution U(a, b), the mean is (a + b)/2 and the variance is (b – a)²/12. A slip occurs when candidates forget that the probability of an interval is proportional to its length, e.g., P(c < X < d) = (d - c)/(b - a), not simply d – c.

对于连续均匀分布 U(a, b)，均值为 (a + b)/2，方差为 (b – a)²/12。考生若忘记区间概率与其长度成正比，例如 P(c < X < d) = (d - c)/(b - a)，而直接写成 d – c，就会出错。

12. Non-parametric Tests: Sign Test and Wilcoxon Signed-Rank | 非参数检验：符号检验与 Wilcoxon 符号秩检验

When the assumption of normality is violated, non-parametric tests become essential. The sign test uses only the direction of differences (positive or negative), discarding magnitude. A common oversight is discarding ties (zero differences) entirely instead of adjusting the effective sample size; ties do not contribute to the test statistic. Candidates then compare the number of positive differences to a binomial distribution with n = number of non-zero differences, p = 0.5.

当正态性假设不成立时，非参数检验变得不可或缺。符号检验仅利用差异的方向（正或负），舍弃大小。常见的疏忽是将零差异（相等）完全丢弃，而未调整有效样本量；零差异不计入检验统计量。考生随后将正差异的个数与二项分布比较，其中 n 为非零差异的个数，p = 0.5。

The Wilcoxon signed-rank test uses both the sign and the magnitude of differences, ranking the absolute differences. Errors stem from misranking tied absolute differences (assign average ranks) and mishandling positive and negative ranks in the test statistic. Also, students may apply the Wilcoxon test to independent samples as if it were a signed-rank test; for two independent samples, the correct procedure is the Wilcoxon rank-sum test (Mann-Whitney U).

Wilcoxon 符号秩检验同时利用差异的符号与大小，对绝对差异进行排序。错误源于对相等绝对差异的秩次处理不当（应分配平均秩次），以及在检验统计量中对正秩与负秩的处理失误。此外，学生可能将 Wilcoxon 检验应用于独立样本，仿佛它是符号秩检验；对于两个独立样本，正确的程序是 Wilcoxon 秩和检验（Mann-Whitney U）。

Choosing between parametric and non-parametric alternatives is a frequent exam decision. Candidates must be prepared to justify their choice based on normality assumptions or the presence of outliers.

在参数与非参数备选方法之间做出选择是考试中的常见决策。考生必须准备好根据正态性假设或异常值的存在性来证明其选择的合理性。

Published by TutorHao | Statistics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

High-Frequency Topics and Common Mistakes in Pre-U OCR Statistics | Pre-U OCR 统计：高频考点与易错题分析