📚 Exam Techniques and Marking Criteria for CIE A-Level Statistics | CIE A-Level 统计答题技巧与评分标准
Success in CIE A-Level Statistics depends on two things that are often taught separately: a secure grasp of statistical ideas, and a strategic approach to putting those ideas on paper in the exact form the examiner is looking for. Every CIE statistics paper is marked with a published scheme that separates method marks, accuracy marks and independent marks, and a candidate who understands that scheme can turn a half-finished attempt into two or three marks. This article explains how marks are awarded in CIE statistics, walks through the main topic areas where technique matters most - probability, discrete distributions, the normal distribution, hypothesis testing and bivariate data - and finishes with practical advice on rounding, notation, calculator use and time management.
在 CIE A-Level 统计中取得好成绩，取决于两件常被分开教授的事：扎实掌握统计概念，以及把概念以考官期待的形式写在答题纸上的策略。CIE 每一份统计试卷都按公开发布的评分方案评分，将方法分、准确性分和独立分区分开来；理解这套方案的考生，可以把一个只写了一半的尝试变成两三分。本文将说明 CIE 统计的给分方式，逐一梳理技巧最重要的板块——概率、离散分布、正态分布、假设检验与双变量数据——最后给出关于精度、记号、计算器使用和时间管理的实用建议。
1. How CIE Statistics Papers Are Marked: M, A and B Marks | CIE 统计试卷如何评分：M、A、B 分
CIE statistics mark schemes use a compact shorthand. A method mark (M1, M2, ...) rewards a correct approach: setting up a hypothesis test in the right form, choosing the correct distribution, substituting into the correct formula, or attempting a valid comparison. A method mark is usually awarded even if the arithmetic afterwards goes wrong, because it is about the route rather than the destination.
CIE 统计评分方案使用一套简洁的缩写。方法分(M1、M2……)奖励正确的解题思路：以正确形式设立假设检验、选择正确的分布、代入正确的公式，或尝试有效的比较。即使之后的算术出错，方法分通常仍会给出，因为它衡量的是路径而不是终点。
An accuracy mark (A1, A2, ...) normally depends on a preceding method mark and requires the final value to be correct to the required degree of accuracy. Independent marks (B1, B2, ...) stand alone: they are given for a correct statement or value that does not depend on any method, for example stating the null and alternative hypotheses correctly, quoting the correct degrees of freedom, or writing down the value of a given probability.
准确性分(A1、A2……)通常依赖于前面的方法分，并要求最终数值精确到规定的小数位或有效数字。独立分(B1、B2……)则独立存在：授予不依赖任何方法的正确陈述或数值，例如正确写出原假设与备择假设、引用正确的自由度，或直接写出题目给定的概率值。
| Code 代码 | Meaning 含义 | Typical example 典型例子 |
| M1 | Method mark 方法分 | Setting up a correct test statistic 写出正确的检验统计量 |
| A1 | Accuracy mark 准确性分 | Correct final value to 3 s.f. 最终值精确到 3 位有效数字 |
| B1 | Independent mark 独立分 | Stating H₀ and H₁ correctly 正确写出 H₀ 与 H₁ |
| ft | Follow-through 追踪误差分 | Method correct, earlier value wrong 方法对、前面数值错 |
| cao | Correct answer only 仅正确答案得分 | No follow-through allowed 不允许追踪误差 |
| oe | Or equivalent 或等价形式 | Equivalent algebraic form accepted 等价的代数形式可接受 |
| soi | Seen or implied 见到或隐含 | Value used but not stated 用到但未明写 |
| sc | Special case 特殊情况 | Alternative valid method 另一种有效方法 |
The practical message is simple: write the setup, do not just write the answer. A correct numerical answer with no visible working may score 1 out of 4, while an incorrect answer with a fully correct route may score 3 out of 4.
实践结论很简单：把过程写出来，不要只写答案。一个没有任何过程的正确答案可能只得 4 分中的 1 分，而一个答案错误但思路完整的解答可能拿到 4 分中的 3 分。
2. Follow-Through, Special Cases and 'Not From Wrong Working' | 追踪误差分、特殊情况分与 nfww
Follow-through (ft) marks exist to protect candidates who make one early slip. If you misread a value in part (a) but use it consistently and correctly in part (b), the later marks are usually still available. The key word is 'consistently': follow-through is only granted when your later work genuinely uses your earlier value, so you must show that value clearly.
追踪误差分(ft)的存在是为了保护在某处早段出错的考生。如果你在(a)问读错了一个数值，但在(b)问中一致且正确地使用了它，后续的分数通常依然可得。关键词是"一致"：只有当你的后续解答确实使用了前面那个错误数值时，追踪误差分才会给出，因此你必须把这个数值清楚地写出来。
Some marks are flagged 'nfww' (not from wrong working), which means the answer must be obtained without relying on a previous error. This appears most often where the question gives a value and asks you to verify or use it, or where a second independent method exists. If you see values supplied in the question stem, treat them as fixed and do not re-derive them unless asked.
有些分数会标注 "nfww"（not from wrong working，即不能依赖前面错误的解答），意思是答案必须在没有沿用先前错误的前提下得出。这种情况最常见于题目已经给出某数值、要求你验证或使用它，或者存在第二种独立解法的场合。如果题干已经给出数值，请把它当作固定值，除非题目要求，否则不要再自行推导。
Special case (sc) marks reward a legitimate alternative method that the main scheme does not cover, such as solving a probability question by listing a sample space instead of using a formula. If your method is valid but unusual, write a short sentence explaining what you are doing so the examiner can see the logic rather than guessing at it.
特殊情况分(sc)奖励评分主方案未涵盖但确实合法的替代方法，例如用列举样本空间代替公式来解概率题。如果你的方法有效但不常见，请用一句话说明你在做什么，让考官看到逻辑，而不是去猜。
3. Reading the Question: Command Words, Context and Mark Tariffs | 审题：指令词、语境与分数分配
Before writing anything, underline the command word. 'Calculate', 'find' and 'work out' need a numerical answer; 'show that' needs a clear derivation ending in the printed value; 'explain', 'comment' and 'interpret' need words, usually referring to the context of the question (the machine, the patients, the packets of sugar); 'state' needs no working at all.
动笔之前，先划出指令词。"Calculate"、"find"、"work out" 需要数值答案；"show that" 需要清晰的推导并以题目给出的数值结尾；"explain"、"comment"、"interpret" 需要文字作答，而且通常要联系题目语境（那台机器、那些病人、那些糖包）；"state" 则完全不需要过程。
The mark tariff tells you how much structure is expected. A 2-mark probability question rarely needs a tree diagram; a 5-mark question about a test almost always wants the hypotheses, the statistic, the comparison and the conclusion written as separate steps. If a question part carries 4 marks and you have written one line, you have almost certainly missed something.
分数分配揭示了题目期望的结构。2 分的概率题基本不需要树形图；5 分的检验题几乎总是要求把假设、统计量、比较和结论分成独立的步骤写出来。如果一个小问占 4 分而你只写了一行，那你几乎肯定漏掉了什么。
Finally, identify the context variables. State the distribution in the words of the question: if X is the number of defective components, then your answer should talk about defective components, not about 'successes'. Context statements frequently carry the final B1 or A1 mark.
最后，识别语境变量。要用题目的语言描述分布：如果 X 表示次品数量，那么你的答案就该谈次品，而不是谈"成功次数"。带语境的结论句往往就是最后那个 B1 或 A1 分。
4. Showing Working: How to Bank Every Method Mark | 书写过程：如何稳稳拿下每一个方法分
A well-presented statistics solution reads like a short argument: what you are assuming, what you are calculating, what you obtain, and what it means. Examiners mark what is on the page, not what is in your head, so any step that carries a mark must be visible.
一份呈现良好的统计解答读起来像一段简短的论证：你假设什么、计算什么、得到什么、意味着什么。考官只看纸面上的内容，不看你的脑内活动，因此任何承载分数的步骤都必须写在卷面上。
For any calculation worth more than one mark, write the general formula first, then substitute, then evaluate. This single habit captures the method mark even when the arithmetic fails, and it makes follow-through possible in later parts.
对于分值超过一分的计算，先写通式，再代入，最后求值。这一个习惯能在算术出错时保住方法分，也让后续小问的追踪误差分成为可能。
- Standardise before you compare 比较前先标准化 - write z = (x − μ) / σ with the numbers substituted.
- State the distribution 写出分布 - X ~ B(n, p) or X ~ N(μ, σ²) earns a mark on its own.
- Keep intermediate values 保留中间值 - store them in calculator memory and round only at the end.
- Label every answer 为每个答案做标注 - P(X ≥ 3) = 0.729, not just 0.729.
A useful discipline is to ask yourself after each line: 'could the examiner give me a mark for this line?' If the answer is no, the line is either a repeat or a gap.
一个有用的自律方法是每写完一行都自问："考官能为这一行给分吗？"如果答案是否定的，那这一行要么是重复，要么就是缺口。
5. Accuracy, Rounding, Units and Statistical Notation | 精度、四舍五入、单位与统计记号
Accuracy marks are lost far more often through premature rounding than through genuine misunderstanding. The standard CIE convention is to give non-exact numerical answers to three significant figures unless the question specifies otherwise, but intermediate values must be kept to full calculator precision.
准确性分的丢失，因过早四舍五入而丢的远比因真正理解错误而丢的多。CIE 的标准约定是：除非题目另有规定，非精确的数值答案保留三位有效数字，但中间值必须保留计算器的完整精度。
Final answer → 3 s.f. | Intermediate values → keep at least 6 s.f. or store in memory
Rounding a probability to three decimal places in the middle of a calculation, or rounding a mean before using it in a test statistic, is one of the most common avoidable errors in the whole paper. Conversely, an answer written as 0.6666667 when 0.667 was expected may still be accepted, but an answer written as 0.67 will not be.
在计算中途把概率四舍五入到三位小数，或者在未把均值代入检验统计量之前就取整，是整份试卷中最常见、也最容易避免的错误之一。反过来说，当题目期待 0.667 时写 0.6666667 可能仍被接受，但写成 0.67 则不予给分。
Notation matters too. Use μ for a population mean and x̄ for a sample mean; use σ for a population standard deviation and s for a sample standard deviation; write probabilities as P(X > 4) rather than 'the chance that X is bigger than 4'. Always include units where the variable is a measurement, and never round a probability to more than three decimal places.
记号同样重要。总体均值用 μ，样本均值用 x̄；总体标准差用 σ，样本标准差用 s；概率写成 P(X > 4)，而不要写成"X 大于 4 的可能性"。当变量是测量值时务必带上单位，并且概率的小数位绝不要超过三位。
6. Probability: Venn Diagrams, Tree Diagrams and Conditional Probability | 概率：韦恩图、树形图与条件概率
Probability questions in CIE statistics reward clear representation more than clever algebra. When a question involves two or three overlapping events, draw the Venn diagram and fill in the regions from the inside out, starting with the intersection. When a question involves a sequence of events, draw the tree diagram and write the probability on every branch, even the ones you think you will not need.
CIE 统计的概率题奖励清晰的表示方法，而不是聪明的代数技巧。当题目涉及两三个相互交叠的事件时，画出韦恩图，从最内层开始填写各区域，先填交集。当题目涉及事件序列时，画出树形图，并在每条分支上写出概率，即使你认为用不到的也要写。
Conditional probability is the single most examinable idea here, and it is almost always the final part of the question. Learn the defining relationship and write it down before substituting.
条件概率是这一板块最常考的概念，而且几乎总是题目最后一问。要熟记定义式，并在代入前先写出来。
P(A | B) = P(A ∩ B) ÷ P(B)
For independence, check whether P(A ∩ B) = P(A) × P(B); for mutually exclusive events, check whether P(A ∩ nts are not independent unless one of them has probability zero. A conclusion line such as 'the events are not independent because 0.24 ≠ 0.3 × 0.9' normally carries the mark.
判断独立性时，检验是否有 P(A ∩ B) = P(A) × P(B)；判断互斥时，检验是否有 P(A ∩ B) = 0，并牢记：互斥事件并不独立，除非其中之一的概率为零。像"两事件不独立，因为 0.24 ≠ 0.3 × 0.9"这样的结论句，通常就是得分点。
Where order matters and repetition is allowed, count systematically rather than guessing: nPr for arrangements without repetition, nCr for selections without order, and nʳ when repetition is permitted.
当顺序重要且允许重复时，要系统地计数而不是靠猜：不重复的排列用 nPr，不计顺序的组合用 nCr，允许重复时用 nʳ。
7. Discrete Random Variables: Binomial and Geometric Distributions | 离散随机变量：二项分布与几何分布
Whenever a question describes a fixed number of independent trials with a constant probability of success, name the distribution explicitly. Writing X ~ B(n, p) with the numbers substituted is usually worth a mark and tells the examiner immediately which tables or calculator function you should be using.
只要题目描述的是固定次数的独立试验且成功概率恒定，就要明确写出分布名称。写出代入数值后的 X ~ B(n, p) 通常就值一分，也立刻告诉考官你该用哪张表或哪个计算器功能。
P(X = r) = ⁿCᵣ pʳ (1 − p)ⁿ⁻ʳ | E(X) = np | Var(X) = np(1 − p)
Watch the direction of the inequality. 'At least three' means P(X ≥ 3) = 1 − P(X ≤ 2); 'fewer than five' means P(X ≤ 4); 'more than two' ent explicitly, because the M1 usually sits on that line rather than on the final decimal.
注意不等号的方向。"至少三个"是 P(X ≥ 3) = 1 − P(X ≤ 2)；"少于五个"是 P(X ≤ 4)；"多于两个"是 P(X ≥ 3)。要把互补的式子明确写出来，因为 M1 通常落在那一行，而不是落在最后的数值上。
A geometric distribution models the number of trials up to and including the first success, so its probability mass function and mean look like this:
几何分布刻画的是直到并包含第一次成功在内的试验次数，因此其分布列与均值如下：
P(X = r) = (1 − p)ʳ⁻¹ p | E(X) = 1 ÷ p
te. Marks are awarded for the table, for the products xP(X = x), and for the final totals.
对于一般的离散随机变量，先列出概率表，再计算。分数分别给在概率表、乘积 xP(X = x) 以及最终合计上。
E(X) = Σ x P(X = x) | Var(X) = E(X²) − [E(X)]²
+ 3), apply the standard rules: E(aX + b) = aE(X) + b and Var(aX + b) = a²Var(X), with no constant added to the variance.
用手算时用 Var(X) = E(X²) − [E(X)]²，比用定义式更快、更不易出错。若题目要求 E(2X + 3) 或 Var(2X + 3)，就套用标准规则：E(aX + b) = aE(X) + b，Var(aX + b) = a²Var(X)，方差中不加常数项。
8. The Normal Distribution and Standardisation | 正态分布与标准化
The normal distribution appears in almost every CIE statistics paper, and almost every question on it begins with standardisation. State the distribution, standardise, then use the table or calculator function, then interpret.
正态分布几乎出现在每一份 CIE 统计试卷中，而几乎每一道相关的题目都从标准化开始。写出分布、标准化、查表或用计算器、再解释。
X ~ N(μ, σ²) → z = (x − μ) ÷ σ → z ~ N(0, 1)
n error here is confusing the variance with the standard deviation. If the question says X ~ ide by 3 and not by 9. Reading the notation carefully is worth more than any calculation in this topic.
这里最常见的错误是把方差与标准差弄混。如果题目写 X ~ N(20, 9)，那么 σ² = 9、σ = 3，因此要除以 3 而不是 9。在这个板块，仔细读懂记号比任何计算都更有价值。
esponding z value from the table, then un-standardise using x = μ + zσ. The sketch itself is not usually marked, but it prevents the classic sign error of using the wrong tail.
逆正态问题要求你由概率反推 x 值。先画出曲线、为所求区域涂色、从表中查出对应的 z 值，再用 x = μ + zσ 反标准化。草图本身通常不给分，但它能避免用错尾部的经典符号错误。
h the continuity correction applied to the boundary. For example, P(X ≥ 20) becomes P(Y ≥ 19.5) after the correction.
当题目把正态分布与二项分布结合起来时，不要忘记正态近似的条件：np > 5 且 n(1 − p) > 5，并且要在分界处做连续性修正。例如 P(X ≥ 20) 修正后变为 P(Y ≥ 19.5)。
9. Hypothesis Testing: Structure, Critical Regions and Conclusions | 假设检验：结构、拒绝域与结论
Hypothesis testing is where structure earns the most marks and where unstructured answers lose them fastest. Follow the same five-step template every single time, and you will rarely drop more than one mark.
假设检验是结构最能挣分、而无结构答案最容易丢分的板块。每次都遵循同样的五步模板，你几乎不会丢超过一分。
- Step 1 - Hypotheses 假设: state H₀ and H₁ with the parameter, e.g. H₀: μ = 250, H₁: μ > 250.
- Step 2 - Distribution 分布: write the sampling distribution, e.g. X̄ ~ N(μ, σ² ÷ n).
- Step 3 - Test statistic 检验统计量: standardise with the correct standard error.
- Step 4 - Compare 比较: compare with the critical value or the significance level.
- Step 5 - Conclude 结论: reject or do not reject H₀, then answer the question in context.
z = (x̄ − μ) ÷ (σ ÷ √n)
Two points about language. First, you never 'accept' H₀; you either 'reject' it or 'do not reject' it, because the test has not proved the null hypothesis true. Second, the conclusion must mention the context: 'there is sufficient evidence at the 5% level that the mean mass of the packets has increased' scores the final mark, while 'reject H₀' on its own usually does not.
关于措辞有两点。第一，永远不要说"接受"H₀，只能说"拒绝"或"不拒绝"，因为检验并没有证明原假设为真。第二，结论必须点明语境："在 5% 显著性水平下，有充分证据表明糖包的平均质量增加了"能拿到最后一分，而只写"拒绝 H₀"通常拿不到。
For a t-test with unknown population variance, the statistic and degrees of freedom change, and quoting ν correctly is usually a B mark.
当总体方差未知而使用 t 检验时，统计量与自由度都会改变，正确写出 ν 通常是一个 B 分。
t = (x̄ − μ) ÷ (s ÷ √n) with ν = n − 1
For a paired-sample test, first form the differences and treat them as a single sample; for a goodness-of-fit or contingency test, compute the expected frequencies before the statistic
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