📚 Common Mistakes in OxfordAQA International A-Level Further Mathematics 9665 Statistics | OxfordAQA A-Level进阶数学9665统计学 易错点总结
In OxfordAQA International A-Level Further Mathematics (9665), the Statistics component challenges students with advanced topics such as continuous random variables, bivariate distributions, hypothesis testing, and non-parametric tests. Many candidates lose marks not because they lack understanding, but because they fall into predictable traps. This article compiles the most frequent errors observed in topic tests and past papers, helping you tighten up your exam technique and secure higher marks. Each pitfall is paired with actionable advice to turn common mistakes into strengths.
在OxfordAQA国际A-Level进阶数学(9665)中，统计学模块涵盖了连续随机变量、二元分布、假设检验和非参数检验等进阶主题。很多考生丢分并非因为理解不透彻，而是掉入了可预见的陷阱。本文整理了单元测试和历年真题中最高频的错误，帮助你优化考试技巧、稳固提升分数。每个易错点都配有实操性建议，让常见错误变成你的得分利器。
1. Confusing Probability Density Function (PDF) with Cumulative Distribution Function (CDF) | 混淆概率密度函数(PDF)与累积分布函数(CDF)
under the curve gives probability. Direct substitution into f(x) only gives the density height, not Pr(X = x), which is zero for continuous variables. Many students incorrectly write P(a < X < b) = f(b) - f(a), which is wrong; this should be F(b) - F(a) where F is the CDF.
经典错误：题目要求计算某个区间内的概率时，学生直接代入概率密度函数(PDF)求值，或者用CDF的方法反过来用在PDF上。PDF f(x)并不是概率，曲线下的面积才是概率。直接代入f(x)得到的只是密度的高度，而不是概率——连续随机变量取单点的概率为零。很多学生错误地写出P(a < X < b) = f(b) - f(a)，但正确做法应该是用累积分布函数F(x)计算：F(b) - F(a)。
Remedy: Always check whether you need to integrate the PDF or differentiate the CDF. For P(a < X < b), integrate f(x) from a to b. If given F(x), simply evaluate F(b) - F(a). And remember: for a valid PDF, the total area must equal 1, so use that to find unknown constants.
解决之道：始终确认是需要对PDF积分还是对CDF求导。计算P(a < X < b)时，若给定PDF f(x)，则从a到b积分；若给定CDF F(x)，则直接计算F(b) - F(a)。还要记住：合法的PDF全域积分必须等于1，可据此求出未知常数。
2. Mishandling the Modulus in Integration for PDFs Defined Piecewise | 分段PDF积分时忽略绝对值或定义域
rong interval or forgetting to split the integral when the domain includes a turning point or a break. Sometimes students apply the modulus incorrectly when a function like f(x) = ½ e-|x| is given, forgetting that |x| changes behaviour at zero.
很多概率密度函数是分段定义的，比如f(x) = kx (0 ≤ x ≤ 2)，其他地方为0。常见错误是积分区间选错，或者在定义域包含转折点或分段点时忘记拆开积分。有时题目给出f(x) = ½ e-|x|这样的函数，学生错误地处理绝对值，忽略了|x|在零点两侧的不同表达式。
d exponential distributions, integrate from -∞ to 0 using -x and from 0 to ∞ using x. Always state the piecewise form before integrating, and watch out for symmetrical properties to halve the workload.
当PDF含有绝对值时，应在绝对值内部表达式变号处拆分积分。双边指数分布要从-∞到0用-x，0到∞用x。积分前务必先写出分段表达式，并留意对称性质可以减半计算量。
3. Misinterpreting Independence and Conditional Probability for Bivariate Distributions | 二元分布中误判独立性与条件概率
Given a joint probability mass function or joint PDF, students often assume independence without verifying fX,Y(x,y) = fX(x)fY xam questions, the variables are deliberately dependent. A frequent mistake is writing P(X| old, or using the formula for covariance as if variables were independent, leading to wrong E(XY) calculations.
给定联合分布列或联合PDF时，学生经常不经验证就假设随机变量独立，实际上题目往往故意设置相关变量。常见错误是在不独立的情况下写出P(X|Y) = P(X)，或者在计算协方差时误用独立性条件下的E(XY)公式，导致错误结果。
Always check independence by testing fX,Y ly using the joint distribution: sum or integrate x*y*f(x,y). For covariance Cov(X,Y) = E(XY) - E(X)E(Y), do not skip the joint calculation.
务必检验独立性：对比联合函数与边缘函数乘积，离散情况检查P(X=x, Y=y)是否等于P(X=x)P(Y=y)。如果不独立，要直接用联合分布计算E(XY)——求和或积分x*y*f(x,y)。协方差公式Cov(X,Y) = E(XY) - E(X)E(Y)中，E(XY)必须用联合分布算，不能偷懒。
4. Errors in Variance and Expectation of Linear Combinations | 线性组合的方差与期望计算错误
ny simply add the variances, ignoring the covariance or assuming independence without justification. Even when variables are independent, stating this explicitly is necessary for exam credit. Also, the variance of a constant times a variable: Var(3X) = 9Var(X), but students sometimes mistakenly multiply by 3 instead of 9.
考生经常在计算Var(aX + bY)时漏掉协方差项。正确公式为a²Var(X) + b²Var(Y) + 2ab Cov(X,Y)。很多人直接加总方差，忽略了协方差，或在未证明独立的情况下默认独立。即使题目条件独立，也必须明确注明才能得分。另外，常数倍变量的方差：Var(3X) = 9Var(X)，却有学生错误地只乘3而不是9。
ch for non-linear transformations: E(X²) is not (E(X))². Use the definition E(X²) = ∫ x² f(x) dx, not the square of the mean.
期望的线性组合E(aX + bY) = aE(X) + bE(Y)始终成立。但E(XY)除非独立或协方差为零，否则不能拆成E(X)E(Y)。务必注意非线性变换：E(X²) ≠ (E(X))²，必须通过∫ x² f(x) dx计算，绝不能简单取均值的平方。
5. Confusing One-Tailed and Two-Tailed Hypothesis Tests | 单尾与双尾假设检验混淆
oard. When setting up H₁, students mis-match the wording of the question. For phrases like 'test whether there is a change', 'different from', or 'has an effect', a two-tailed test is required. For 'increase', 'greater than', 'improve' iled significance level (e.g., splitting α = 0.05 into 0.025 for each tail but using the 5% ne-tailed table). Double-check the rejection region and halving of p-values appropriately.
这是失分重灾区。设定备择假设H₁时，学生经常误判题目措辞。出现'检验是否有变化'、'不等于'、'是否有影响'等词，应使用双尾检验；而'增加'、'大于'、'改善'等对应单尾检验。另一个典型错误：临界值用单尾，而显著性水平却按双尾拆分（比如把α=0.05每尾分0.025，但查表取了单尾5%临界值）。务必核对拒绝域和对p值的恰当处理。
A reliable method: Write H₀ and H₁ clearly with strict inequality signs. For two-tailed, H₁ has '≠' and p-value is doubled (or compared with α/2 if using critical regions). For one-tailed, H₁ has '<' or '>' and the p-value is not doubled. Never decide the tails after seeing the data.
可靠方法：严格用不等式写出H₀和H₁。双尾检验H₁含'≠'，p值加倍（或用α/2与临界值比较）。单尾检验H₁含'<'或'>'，p值不变。切勿在看到数据之后再决定单双尾。
6. Chi-Squared Tests: Degrees of Freedom and Validity Conditions | 卡方检验：自由度与有效性条件
m ν = (r-1)(c-1). Many students mistakenly use r×c - 1. For goodness-of-fit, ν = number of categories - 1 - number of estimated parameters. Also, the approximation to χ² distribution is poor if any expected frequency is less than 5; the Yates' ing observed frequencies instead of expected in the test statistic formula Σ (O - E)² / E.
独立性检验和拟合优度卡方检验有特殊陷阱。r行c列的列联表，自由度ν = (r-1)(c-1)，很多学生错用r×c - 1。拟合优度检验中，ν = 类别数 - 1 - 估计的参数个数。此外，当任一期望频数小于5时，卡方近似不佳；2×2表可使用叶茨修正，或合并类别。另一个典型错误：在统计量公式Σ (O - E)² / E中误用观测频数代替期望频数。
he null hypothesis. For independence, E = (row total × column total) / grand total. If >20% of expected cells are below 5, consider Fisher' for each parameter estimated from the data, such as the mean in a Poisson goodness-of-fit.
始终基于原假设计算期望频数。独立性检验中，E = (行合计 × 列合计) / 总合计。如果低于5的期望格值占比超过20%，应考虑费希尔精确检验或合理合并行/列。还要记住，当从数据估计参数（如泊松拟合优度中的均值）时，每估计一个参数自由度减1。
7. Errors in Confidence Intervals for Mean and Proportion | 均值和比例置信区间的错误
; use t when variance is estimated from the sample, with ν = n - 1. However, if n is large (n > 30), many exam boards accept z as an approximation to t. But for A-Level Further Maths, precise use of t-distribution is expected for small samples from a normal population. Another mistake: forgetting to check the underlying assumption of normality for small samples, which is required for the t-interval to be valid.
构建置信区间时，学生常因选错z值或t值而出错。当总体方差已知时用z分布；当方差由样本估计时，用t分布，自由度ν = n - 1。不过当样本量较大(n > 30)时，很多考试局允许用z近似t，但在进阶数学中，小样本且正态总体时要求严格使用t分布。另一个错误是：忘记检验小样本的正态性假设，t区间在该假设下才有效。
Write the correct generic formula for mean: CI = x̄ ± tn-1, α/2 * s/√n (or z). For proportion: p̂ ± z * √[p̂(1-p̂)/n]. Check the success-failure condition: np̂ ≥ 10 and n(1-p̂) ≥ 10 for the normal approximation. Never use t with proportions.
正确写出均值置信区间公式：x̄ ± tn-1, α/2 * s/√n（或z）。比例区间：p̂ ± z * √[p̂(1-p̂)/n]。检查成功-失败条件：np̂ ≥ 10且n(1-p̂) ≥ 10才能用正态近似。在比例问题中绝不能使用t分布。
8. Mishandling Continuous Random Variables: Median, Quartiles, and Percentiles via CDF | 连续随机变量的中位数、四分位数和百分位数计算错误
1) = 0.25, upper quartile F(Q3) = 0.75. When the piecewise CDF crosses 0.5 in a different piece, they solve the wrong branch. Always identify which segment of the CDF corresponds to the required probability and ensure the solution lies within that interval.
求中位数m时，应解F(m) = 0.5，而不是f(m) = 0.5。令人惊讶的是大量学生直接把0.5代入PDF。同样，下四分位数F(Q1)=0.25，上四分位数F(Q3)=0.75。当分段CDF在某个片段穿过0.5时，学生经常解错分支。务必先判断所需概率落在CDF的哪一段，确保解在该区间内。
in the correct domain. And don’t forget that for continuous symmetric distributions, median = mean, but this is not a universal shortcut.
良好习惯：画出CDF草图，从0单调递增到1。估计0.5穿过的大致位置。求解后进行区间验证。注意连续对称分布的中位数=均值，但并非普遍成立，勿滥用。
9. Errors with Correlation and Regression: Extrapolation and Reversing Variables | 相关与回归中的错误：外推与变量反转
, especially if the relationship is not linear beyond the data range. The exam expects you to comment on the danger of extrapolation. Another common mistake: interchanging the regression line of y on x with that of x on y. The coefficient b in y = a + bx is calculated as Sxy / Sxx, whereas for x on y the slope is Sxy / Syy. Mixing these up leads to completely wrong equations.
学生常拿回归线y = a + bx去预测原始x数据范围以外的值（外推），这可能会得出荒谬的结果，尤其在数据范围外关系并非线性的情况下。考试期望你评价外推的风险。另一个常见错误：混淆y对x的回归线与x对y的回归线。y = a + bx中的系数b = Sxy / Sxx，而x对y的回归斜率是Sxy / Syy。混淆两者会导致完全错误的方程。
s are only valid within the range of observed x-values. Also, correlation coefficient r = Sxy / √Sxx Syy is symmetric, but the regression lines are not.
给定汇总统计量时，先确定因变量。用x预测y时，应使用y对x的回归线。明确说明预测仅在观测x范围内有效。另外，相关系数r = Sxy / √Sxx Syy是对称的，但两条回归线并不对称。
10. Misapplication of the Poisson Approximation to Binomial or Other Distributions | 泊松逼近二项分布或其他分布的误用
p is not small (e.g., p = 0.4) or when n is moderate but np is large. Also, applying a continuity correction incorrectly can cause errors; however, for Poisson approximation, continuity correction is not applied as strictly as in normal approximation. But when moving from Poisson to normal (λ > 10), a continuity correction is expected. Many candidates forget to add or subtract 0.5 appropriately.
二项分布Binomial(n, p)近似为Poisson(np)的条件是n大、p小（一般np < 10，典型要求n > 50且np < 5）。学生常在p不小时（如p=0.4）或np较大时误用该近似。另外，连续性修正使用不当也会出错；泊松近似时通常不强求连续性修正，但从泊松近似到正态(λ > 10)时则需要进行连续性修正。很多学生忘记适当加减0.5。
λ, variance = λ, and apply continuity correction: P(X ≤ k) ≈ P(Z < (k + 0.5 - λ)/√λ). Don't correct for exact values in Poisson if not required.
务必检查条件：二项到泊松，安全规则n > 50, p < 0.1。泊松到正态，λ > 10，均值λ，方差λ，加上连续性修正：P(X ≤ k) ≈ P(Z < (k + 0.5 - λ)/√λ)。泊松分布本身求精确概率时通常不要求修正。
11. Incorrect Use of Central Limit Theorem and Standard Error | 中心极限定理与标准误使用不当
When dealing with sample means, the standard error is σ/√n or s/√n e sample mean is approximately normally distributed regardless of the population distribution. But many students forget to check whether n is large enough (usually n ≥ 30) before applying the CLT. Furthermore, if the population is already normal, the sample mean is exactly normal for any n, and no CLT is needed.
处理样本均值时，标准误是σ/√n或s/√n。常见错误是直接把总体标准差当作样本均值的标准差。中心极限定理(CLT)指出，对大样本，无论总体分布如何，样本均值近似正态。但很多学生在应用CLT前忘记检查n是否足够大（通常n ≥ 30）。另外，若总体本身正态，样本均值对于任何n都是精确正态，无需引用CLT。
ard deviation. In hypothesis tests for a mean with unknown σ, the test statistic is t = (x̄ - μ) / (s/√n). Never write denominator as just s or σ.
构建均值检验统计量或置信区间时，分母用标准误，而非原始标准差。未知σ的均值假设检验，统计量为t = (x̄ - μ) / (s/√n)。分母绝不能写成单纯的s或σ。
12. Forgetting to Check Conditions for Parametric Tests and State Conclusions in Context | 忽略参数检验条件及未将结论置于题目语境
Many marks are lost on hypothesis test conclusions that are not expressed in context. Writing 'reject H₀' is insufficient; you must state: 'There is sufficient evidence at the 5% significance level to suggest that the mean IQ has increased'. Similarly, failing to mention assumptions like normality for t-tests or equal variances for two-sample tests can cost marks. In non-parametric tests like sign test or Wilcoxon signed-rank test, the assumption is often that the data are paired or differences are symmetric, not that the data are normal — mixing these up leads to wrong test choice.
假设检验结论不结合语境丢分严重。"拒绝H₀"这样写是不够的；必须写出："在5%显著性水平下，有足够证据表明平均智商有所提高"。同样，未提及t检验的正态性假设，或双样本检验的方差齐性假设，会损失分数。在非参数检验如符号检验或威尔科克森符号秩检验中，常要求数据配对或差值对称，而不是正态性——混淆这些导致选错检验方法。
Develop a checklist for each test: (1) State H₀ and H₁ with symbols. (2) Identify test statistic. (3) State assumptions. (4) Compute value. (5) Compare with critical value or find p-value. (6) Contextualized conclusion. Always use the wording from the question when phrasing the conclusion.
为每个检验建立清单：(1) 用符号陈述H₀和H₁。(2) 确定检验统计量。(3) 说明假设条件。(4) 计算统计量值。(5) 与临界值比较或求p值。(6) 语境化结论。写结论时尽量使用题目中的措辞。
Published by TutorHao | Further Mathematics 9665 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