AQA A-Level数学S2高分范文详解 | AQA A-Level Maths S2: High-Scoring Model Answers Explained

AQA A-Level数学S2高分范文详解

AQA A-Level Mathematics S2 (MA04) 是统计学模块中最具挑战性的单元之一。它不仅要求你熟练掌握泊松分布、连续随机变量、卡方检验和置信区间等核心知识点，更考验你在限时考试中将理论应用于具体题目场景的能力。很多同学知识点都懂，但一进考场就卡在"怎么写才能拿满分"这道坎上。本文精选四道S2典型例题，逐题展示高分范文的完整推导过程，并标注每道题的评分要点（Mark Scheme关键得分点），让你看清AQA阅卷官到底在找什么。建议先自己动手做一遍例题，再对照范文找出差距——这种对比式学习是短期内提分最快的方式。

AQA A-Level Mathematics S2 (MA04) is one of the most demanding statistics modules. It requires not only mastery of core topics such as the Poisson distribution, continuous random variables, chi-squared tests, and confidence intervals, but also the ability to apply theory to specific problem contexts under timed exam conditions. Many students understand the concepts but freeze when it comes to writing model-quality answers that earn full marks. This article presents four carefully selected S2 exam questions with complete model answer walkthroughs, annotated with mark scheme key points so you understand exactly what AQA examiners are looking for. We recommend attempting each question yourself before comparing with the model answers — this comparative approach is the fastest way to improve your score.

一、泊松分布假设检验的完整推导过程 | Poisson Distribution Hypothesis Testing: Complete Derivation

泊松分布是S2考试中出现频率最高的分布之一。AQA特别喜欢考察泊松可加性（additivity of independent Poisson variables）这一性质——即多个独立泊松变量的总和仍然服从泊松分布，参数为各参数之和。这个性质看似简单，但在假设检验中正确使用它来汇总多时段或多样本数据，是拿分与丢分的分水岭。下面的例题正是2019年AQA真题的典型变体，展示了泊松假设检验的完整评分链路。

The Poisson distribution is one of the most frequently tested distributions in S2. AQA particularly likes to examine the additivity property — that the sum of independent Poisson variables is itself Poisson with the parameters summed. This property seems simple on the surface, but using it correctly in hypothesis testing to aggregate multi-period or multi-sample data is where marks are won or lost. The following example, a typical variant of a 2019 AQA past paper question, demonstrates the complete scoring chain for a Poisson hypothesis test.

1.1 泊松检验的题目分析 | Question Breakdown

Question: A company claims that the number of defective items produced per hour follows a Poisson distribution with mean 2.5. A quality control inspector records the number of defectives per hour over a random sample of 10 hours and obtains a total of 40 defectives. Test, at the 5% significance level, whether there is evidence that the mean number of defectives per hour has increased.

中文题目：某公司声称其每小时次品数量服从均值为 2.5 的泊松分布。质检员随机抽取 10 小时的生产记录，总共发现 40 个次品。在 5% 显著性水平下，检验是否有证据表明每小时的平均次品数增加了。

读题关键点：注意两个关键信号——① "Poisson distribution with mean 2.5" 提示原假设参数 λ = 2.5；② "has increased" 明确指示这是单尾上尾检验（upper-tail test），备择假设为 λ > 2.5。另外，样本数据是10个小时的总次品数 40，不是每小时均值——这意味着你必须先用泊松可加性将10个 Po(2.5) 合并为 Po(25)，再对总观测值 40 做检验。直接对单小时均值 4 做Po(2.5)的检验是致命错误，AQA会在M1步骤直接扣分。

1.2 泊松检验高分范文 | Model Answer with Mark Annotations

Step 1 — Define hypotheses and distribution（定义假设与分布，B1分）

Let X be the number of defectives in a one-hour period. Then X ~ Po(λ).

H₀: λ = 2.5 — the mean number of defectives per hour has not changed.
H₁: λ > 2.5 — the mean number of defectives per hour has increased (upper-tail test).

B1评分标准：必须同时给出H₀和H₁，且明确标注参数（λ而非笼统的"mean"）。H₁必须体现上尾方向——写"λ ≠ 2.5"会被判定为方向错误，即使后续计算正确也拿不到这道题的满分。题目中"has increased"是唯一的正确信号词，不要自己发挥。

Step 2 — Apply Poisson additivity to pool observations（汇总观测值，M1分）

Since X₁, X₂, ..., X₁₀ are independent Po(2.5) random variables under H₀, the total number of defectives over 10 hours follows:

T = ΣXᵢ ~ Po(10 × 2.5) = Po(25)

这10个随机变量相互独立（不同小时的次品数互不影响），因此总和仍为泊松分布，参数直接相加。这是泊松分布区别于二项分布的一个重要特性——当你面对多个时段的计数数据时，泊松汇总计算最为简洁。

⚠️ M1评分要点：必须明确写出"independent"和"Po(25)"。只写计算结果不写"independent Poisson → Po(25)"的推理链，会被视为论证不完整而扣分。

Step 3 — Calculate the p-value（计算p值，M1 + A1分）

Observed total: t = 40. Under H₀, T ~ Po(25).

p-value = P(T ≥ 40 | λ = 25) = 1 − P(T ≤ 39)

Using Poisson cumulative distribution tables or calculator:

P(T ≤ 39) = Σ_{k=0}^{39} (e⁻²⁵ × 25ᵏ / k!) ≈ 0.9963

Therefore: p-value ≈ 1 − 0.9963 = 0.0037

M1A1评分标准：P(T ≥ 40) = 1 − P(T ≤ 39)的转化正确得M1（注意是≤39而非≤40，因为是离散分布）。查表结果0.0037得A1。如果写成P(T ≥ 40)直接用表格查但查错了，M1仍在但A1丢失。

Step 4 — Compare with significance level and conclude（比较与结论，A1分）

Since p-value = 0.0037 < 0.05, we reject H₀ at the 5% significance level.

There is sufficient evidence to suggest that the mean number of defective items produced per hour has increased beyond 2.5. The company's claim is not supported by the sample data.

中文结论：p值0.0037远小于显著性水平0.05，我们在5%水平下拒绝原假设。有充分统计证据表明，每小时的平均次品数已经超过2.5——公司声称的生产标准在样本数据中未得到支持。

A1评分标准：结论必须包含三个要素：① 显著性水平（5%）；② 上下文解释（"defective items per hour has increased"）；③ 明确的reject/do not reject声明。三个要素缺一不可，任何一个遗漏都会扣掉A1。这是AQA全系列统计题目的通用评分模板。

1.3 泊松检验三大常见错误 | Top 3 Poisson Testing Mistakes

❌ 错误1 — 忽略了泊松可加性：直接用单小时均值40÷10=4代入检验，写P(Po(2.5) ≥ 4)。这是AQA阅卷报告中年年出现的经典错误。泊松分布的检验对象必须是可加总后的总分布，因为单个小时的数据信息量不足以支撑统计推断。正确做法永远是先汇总，再检验。

❌ 错误2 — 检验方向判断错误：写成双尾检验（λ ≠ 2.5）。题目中"has increased"是英语中最明确的上尾信号之一。只有出现"has changed"、"is different"或纯提问是否变化时才考虑双尾。方向一旦写错，即使后续全部正确也最多拿一半分数。

❌ 错误3 — p值计算中离散分布边界处理不当：对于离散分布，P(X ≥ 40) ≠ 1 − P(X ≤ 40)，正确写法是1 − P(X ≤ 39)。差这一个下标就可能改变p值，尤其当40本身有不可忽略的概率质量时。这个问题在泊松和卡方检验中都会出现，务必注意。

二、连续随机变量PDF与CDF的互推运算 | Converting Between PDF and CDF for Continuous Random Variables

连续随机变量的概率密度函数（PDF）与累积分布函数（CDF）之间的互推是S2中的必考内容，通常会占8-10分。这类题目的特点是步骤清晰、得分点明确——只要你的积分计算不出错，这是最容易拿满分的题型。AQA的标准命题形式是给你一个含未知常数k的分段PDF，然后让你依次完成：验证k值 → 推导CDF → 计算特定区间的概率。三个子问题环环相扣，前一步算错会连锁影响后面，所以每一步都要仔细验证。

Continuous random variables and the conversion between probability density functions (PDFs) and cumulative distribution functions (CDFs) are guaranteed to appear in S2, typically worth 8-10 marks. These questions have a clear, predictable structure with well-defined marking points — as long as your integration is correct, they are the easiest way to secure full marks. The standard AQA format gives you a piecewise PDF with an unknown constant k, then asks you to: verify k → derive the CDF → calculate a specific probability. The three sub-questions are interdependent — an error in the first step cascades through the rest, so verify each step carefully.

2.1 PDF与CDF的题目分析 | Question Breakdown

Question: A continuous random variable X has probability density function f(x) where:

f(x) = { kx²(3 − x),  0 ≤ x ≤ 3
        { 0,                      otherwise

(a) Show that k = 4/27.
(b) Find the cumulative distribution function F(x).
(c) Calculate P(1 < X < 2).

中文题目：连续随机变量X的概率密度函数如上所示。(a) 证明k = 4/27；(b) 求累积分布函数F(x)；(c) 计算P(1 < X < 2)。

题目结构分析：这是一道典型的PDF三连问。(a)问利用概率总和为1的条件建立积分方程，解出k——纯积分题，2-3分；(b)问考察F(x) = ∫_{-∞}^x f(t) dt的概念理解，需要分段定义，3-4分；(c)问可以直接用(b)的结果F(2)−F(1)，也可以直接积分PDF，无论哪种方法结果一致但用CDF更快且不易出错，2分。整题约8分，在S2试卷中属于中等偏易的基础题。

2.2 PDF-CDF互推高分范文 | Model Answer

(a) Verify k = 4/27 using the total probability condition（M1 + A1）

For any valid PDF, ∫_{-∞}^{∞} f(x) dx = 1:

∫₀³ kx²(3 − x) dx = k ∫₀³ (3x² − x³) dx

= k [x³ − x⁴/4]₀³ = k [(27 − 81/4) − 0] = k × (27/4) = 1

⇒ k = 4/27 ✓

M1评分点：写下∫₀³ f(x) dx = 1即得方法分。A1评分点：正确计算出k = 4/27。中间积分展开步骤可简略但不能完全跳过——至少展示∫₀³(3x²−x³)dx这一步。

(b) Derive the CDF — three-region piecewise definition（B1 + M1 + A1）

F(x) = P(X ≤ x) = ∫_{-∞}^x f(t) dt. Three regions:

Case 1: x < 0 → F(x) = 0（PDF在负区间为零，累积概率自然为0）

Case 2: 0 ≤ x ≤ 3 → F(x) = ∫₀^x (4/27)t²(3 − t) dt

= (4/27) ∫₀^x (3t² − t³) dt = (4/27) [t³ − t⁴/4]₀^x

= (4/27)(x³ − x⁴/4) = x³(4 − x) / 27

Case 3: x > 3 → F(x) = 1（PDF在x>3区间积分为0，全部概率已累积完毕）

Final CDF: F(x) = { 0,                 x < 0
                    { x³(4−x)/27,  0 ≤ x ≤ 3
                    { 1,                x > 3

评分要点：三个区间的完整分段定义各得B1。其中x > 3时F(x) = 1是最容易被遗漏的——很多考生只写了0 ≤ x ≤ 3积出来后就直接结束，少写x > 3的情况会被扣A1。AQA阅卷记录显示约有30%的考生在这里丢分。

(c) Calculate P(1 < X < 2) — two methods（M1 + A1）

Method 1（推荐 — 用CDF）：P(1 < X < 2) = F(2) − F(1)

F(2) = 2³(4 − 2)/27 = 8 × 2/27 = 16/27
F(1) = 1³(4 − 1)/27 = 3/27 = 1/9

P(1 < X < 2) = 16/27 − 1/9 = 16/27 − 3/27 = 13/27 ≈ 0.4815

Method 2（直接积分验证）：∫₁² (4/27)x²(3−x)dx = (4/27)[x³−x⁴/4]₁² = (4/27)(4−0.75) = 13/27 ✓

M1评分点：写出F(2)−F(1)或∫₁² f(x) dx均可得方法分。A1评分点：最终结果13/27，保留分数形式是最佳做法（AQA接受精确分数或≥3 d.p.的小数）。两种方法结果一致可互相验证。

2.3 PDF-CDF互推常见错误 | Common Integration Mistakes

❌ 错误1：求F(x)时积分上限写成x=3而非变量x。F(x)是x的函数，积分上限必须是变量x而非固定值3——这是理解CDF概念的核心。写∫₀³(4/27)t²(3−t)dt得到的F(3)=1，但那是定值不是函数。

❌ 错误2：忘记写x>3时的F(x)=1分段，或写成0≤x≤3时的表达式在x>3时也适用。f(x)在(3,∞)为0，所以x>3时F(x)恒为1而非继续增长。

❌ 错误3：积分展开时符号错误——(x³ − x⁴/4)代入上限时x⁴/4的负号容易遗漏，尤其在用计算器快速算时更需要手工核对。

三、卡方独立性检验列联表完整计算 | Chi-Squared Independence Test: Complete Contingency Table

卡方独立性检验（Chi-squared test for independence）在S2中通常以表格题形式出现，分值在10-12分之间，是分值最重的单个题目类型之一。它的难度不在于单个步骤的复杂性，而在于步骤繁多——从假设陈述、期望频数计算、χ²统计量汇总、自由度确定到最终结论，任何一个环节的疏漏都会丢分。更关键的是，AQA对χ²检验的语言表述有非常严格的规范，下文会详细说明。

The chi-squared test for independence typically appears as a table-based question in S2, worth 10-12 marks — making it one of the heaviest single question types. Its difficulty lies not in any individual step's complexity, but in the sheer number of steps: from hypothesis statement, expected frequency calculation, χ² statistic summation, degrees of freedom determination, to the final conclusion — a mistake at any stage costs marks. More critically, AQA enforces very strict language conventions for χ² conclusions, detailed below.

3.1 卡方检验的题目分析 | Question Breakdown

Question: A researcher surveys 200 students to investigate whether there is an association between gender and preferred learning style (Visual, Auditory, or Kinesthetic). The observed frequencies are:

Visual Auditory Kinesthetic Total Male 45 30 25 100 Female 35 50 15 100 Total 80 80 40 200

Test, at the 1% significance level, whether gender and learning style are independent.

中文题目：研究者调查200名学生，探究性别与学习风格（视觉型/听觉型/动觉型）之间是否存在关联。在1%显著性水平下进行卡方独立性检验。

题目结构分析：2行×3列列联表，总样本量200。自由度ν = (2−1)(3−1) = 2，临界值χ²₂(0.01) = 9.210。这道题的关键陷阱在于显著性水平是1%而非常见的5%——临界值显著变大，增加了不拒绝H₀的可能性。数据本身（男性偏向视觉、女性偏向听觉）看起来有差异，但1%的严格门槛可能导致结论反转——这正是AQA考察"statistical vs practical significance"的经典切入点。

3.2 卡方检验高分范文 | Model Answer

Step 1 — State hypotheses（B1分）

H₀: Gender and learning style are independent — there is no association between the two variables.
H₁: Gender and learning style are not independent — an association exists.

B1评分标准：卡方独立性检验的假设必须用"independent / not independent"的标准措辞，不可写成"no difference"等模糊表述。这是AQA评分指南中的硬性规定。

Step 2 — Calculate expected frequencies（M1分）

Eᵢⱼ = (Row Total × Column Total) / Grand Total

E₁₁(Male-Visual) = 100×80/200 = 40   E₁₂(Male-Auditory) = 100×80/200 = 40
E₁₃(Male-Kinesthetic) = 100×40/200 = 20  E₂₁(Female-Visual) = 100×80/200 = 40
E₂₂(Female-Auditory) = 100×80/200 = 40   E₂₃(Female-Kinesthetic) = 100×40/200 = 20

M1评分标准：只需正确计算至少两个期望频数即得方法分。但建议全部写出——如果计算有误，至少两个正确就能保住M1。

Step 3 — Calculate the χ² test statistic（M1 + A1分）

χ² = Σ (O − E)² / E

= (45−40)²/40 + (30−40)²/40 + (25−20)²/20 + (35−40)²/40 + (50−40)²/40 + (15−20)²/20

= 25/40 + 100/40 + 25/20 + 25/40 + 100/40 + 25/20

= 0.625 + 2.5 + 1.25 + 0.625 + 2.5 + 1.25 = 8.75

M1A1评分标准：公式代入正确得M1，最终值8.75得A1。六个项的加总必须完整写出——跳过中间步骤只给最终答案的，AQA通常会要求展示至少部分计算过程。

Step 4 — Degrees of freedom and critical value（B1分）

ν = (r − 1)(c − 1) = (2 − 1)(3 − 1) = 2

At 1% significance level: χ²₂(0.01) = 9.210 (from statistical tables)

B1评分标准：自由度的公式和数值必须都写出。临界值查对表得B1——注意1%的临界值9.210远大于5%的5.991，这是本题区分度所在。

Step 5 — Conclusion（A1分）

Since χ² = 8.75 < 9.210, we do not reject H₀ at the 1% significance level.

There is insufficient evidence to suggest an association between gender and learning style. Although the observed data shows some differences in distribution (males appear more visual, females more auditory), these differences are not statistically significant at the stringent 1% level.

中文结论：χ² = 8.75 小于临界值 9.210，在1%显著性水平下不拒绝H₀。没有充分证据表明性别与学习风格之间存在关联。虽然观察数据中存在分布差异（男性倾向视觉型、女性倾向听觉型），但在严格的1%水平下这些差异不具统计显著性。

A1评分标准：结论必须：① 比较χ²与临界值（或p值与α）；② 明确"do not reject"（绝不写"accept"）；③ 在题目背景中解释结论含义。三个要素全部满足得A1。特别强调——即使你个人认为数据"看起来有关联"，统计结论必须以计算为准。

3.3 卡方检验五大常见失分点 | Top 5 Chi-Squared Mistakes

❌ 错误1：在2×3及更大的列联表中错误使用Yate's连续性校正。Yate's correction仅适用于2×2列联表，任何大于2×2的表都不使用。

❌ 错误2：期望频数四舍五入导致累积误差。例如将期望值40.5保留为41，最终χ²可能偏离0.1-0.2，虽不改变结论但偶尔会跨越临界值边界。

❌ 错误3：写"accept H₀"。这在AQA评分标准中是明确的扣分项——只能说"do not reject H₀"。不拒绝≠接受，这是统计学的基本理念。

❌ 错误4：合并单元格后忘记调整自由度。如果某期望频数<5而合并了单元格，自由度必须相应减少。

❌ 错误5：结论不引用显著性水平。只说"no association"而不提"at the 1% level"——缺少这半句话直接丢A1。

四、正态分布置信区间构造与样本量反向设计 | Confidence Intervals and Inverse Sample Size Determination

置信区间是S2中与假设检验并重的核心板块。AQA在此题型上的独特考察方式是要求你不仅能正向构建区间，还要能反向计算所需样本量——给定目标区间宽度，倒推出满足精度要求的最小样本量n。这涉及不等式求解和向上取整两个关键操作。更隐蔽的陷阱在于：σ已知时用z分布，σ未知时用t分布——在S2考试中，题目通常会明确给出σ的值或暗示σ已知，此时必须用z而不用t，否则得到的区间会比正确答案更宽。

Confidence intervals are a core S2 topic alongside hypothesis testing. AQA's distinctive approach is to test not only forward interval construction, but also inverse sample size calculation — given a target interval width, derive the minimum n that achieves the required precision. This involves inequality solving and ceiling rounding. A subtler trap: use the z-distribution when σ is known, and the t-distribution when σ is unknown. In S2 exams, the question typically states σ explicitly or implies it is known — you must use z, not t, or your interval will be wider than the correct answer.

4.1 置信区间与样本量题目分析 | Question Breakdown

Question: The heights of a certain species of plant are normally distributed with standard deviation σ = 4.2 cm. A random sample of 16 plants gives a sample mean of 38.5 cm.

(a) Construct a 95% confidence interval for the population mean height μ.
(b) Determine the minimum sample size so that a 95% confidence interval for μ has a width of at most 3 cm.

中文题目：某植物高度服从正态分布，总体标准差σ = 4.2 cm。随机抽取16株测得样本均值38.5 cm。(a) 构建μ的95%置信区间；(b) 求使95%置信区间宽度不超过3 cm所需的最小样本量。

关键判断：σ已知 → 使用z分布。虽然样本量n=16不算大，但题目明确给出σ=4.2（总体标准差），这意味着不需要用样本标准差s来估计，直接用z分布。如果题目只给了样本标准差s，即使n>30也需要考虑t分布。

4.2 置信区间与样本量高分范文 | Model Answer

(a) 95% confidence interval with known σ（M1 + A1）

Formula: x̄ ± z_{α/2} × σ/√n

Given: x̄ = 38.5, σ = 4.2, n = 16, z_{0.025} = 1.96

Standard error: SE = 4.2/√16 = 4.2/4 = 1.05
Margin of error: ME = 1.96 × 1.05 = 2.058

95% CI = (38.5 − 2.058, 38.5 + 2.058) = (36.44, 40.56) cm

Interpretation: We are 95% confident that the true population mean height lies between 36.44 cm and 40.56 cm.

M1A1评分标准：公式代入正确得M1，最终区间(36.44, 40.56)得A1。注意区间必须包含单位cm——不写单位或写错单位直接扣A1。区间表达用括号(36.44, 40.56)而非不等式形式。

(b) Inverse sample size calculation（M1 + A1）

Width of CI = 2 × ME = 2 × z_{α/2} × σ/√n

Require Width ≤ 3 cm:

2 × 1.96 × 4.2/√n ≤ 3 → 16.464/√n ≤ 3 → √n ≥ 5.488 → n ≥ 30.12

Since n must be an integer: n = 31（向上取整，ceiling function）

Verification with n=30: Width = 2×1.96×4.2/√30 = 16.464/5.477 ≈ 3.006 > 3 cm ✗
Verification with n=31: Width = 2×1.96×4.2/√31 = 16.464/5.568 ≈ 2.958 < 3 cm ✓

M1A1评分标准：宽度不等式设置正确得M1，最终n=31得A1。写n=30因为未向上取整会直接丢A1——这是AQA阅卷中的高频扣分点。建议加上验证步骤证明n=31确实是满足条件的最小整数，这样可以拿满B1（如果有额外验证分）。

4.3 置信区间三大常见错误 | Top 3 Confidence Interval Mistakes

❌ 错误1：已知σ时误用t分布，写t_{15}(0.025) ≈ 2.131代替z=1.96。这会得到更宽的区间(36.26, 40.74)——虽然不是严格意义上的"错误"，但在AQA评分中因方法选择不当会被扣M1。题目给了σ就必须用z。

❌ 错误2：混淆"区间宽度"和"误差边际"。Width = 2×Margin of Error。题目要求width≤3意味着ME≤1.5，而非直接代入ME≤3。

❌ 错误3：样本量计算结果向下取整（n=30）而非向上取整（n=31）。这是最常见的低级错误之一——因为30.12>30，n=30不满足条件，"minimum integer that satisfies"一定是31。

以上四道例题覆盖了AQA S2考试中最核心的四大题型。建议将这四篇范文的结构作为你自己的答题模板——假设陈述 → 分布确定 → 统计量计算 → 临界值/p值 → 上下文结论。这个五步框架适用于S2中几乎所有的假设检验和推断类题目。考前用近五年的AQA past papers反复练习这套模板，考试时按部就班写下来，分数自然不会差。

The four examples above cover the four most important question types in AQA S2. We recommend using the model answer structure as your own answer template — hypothesis statement → distribution specification → test statistic calculation → critical value / p-value → contextual conclusion. This five-step framework works for virtually all hypothesis testing and inference questions in S2. Drill this template with recent AQA past papers before the exam and the marks will follow naturally.

需要更多 A-Level 数学辅导？

AQA · Edexcel · OCR · 所有模块一对一精讲

16621398022（同微信）

www.aleveler.com

更多咨询请联系16621398022（同微信）