📚 CIE Statistics Year 12: Key Points for Experimental/Practical Assessment | CIE 统计 Year 12：实验/实践考核要点
A significant part of the CIE AS Level Statistics (Paper 5 of 9709) involves understanding how to plan, conduct, and critique statistical investigations. Students are expected to demonstrate knowledge of experimental design, data collection methods, and the control of bias – all of which form the backbone of practical statistical work. This article distils the essential points for exam success in any question that touches on experimental or practical contexts.
CIE AS 统计（9709 卷五）中有很大一部分内容是理解如何规划、实施和评价统计调查。学生需要展现对实验设计、数据收集方法以及偏差控制的掌握——这些正是实践统计工作的核心。本文提炼了应对实验或实践情境考题的关键要点，助你备考事半功倍。
1. The Nature of Statistical Investigations | 统计调查的本质
Every statistical investigation begins with a clear research question or hypothesis. From an exam perspective, you must be able to distinguish between surveys, observational studies, and designed experiments. A survey gathers information from a sample without manipulating variables, while an experiment deliberately imposes a treatment to measure its effect on a response variable. The choice of method affects the strength of causal conclusions and the presence of confounding factors.
每一项统计调查都始于一个明确的研究问题或假设。从考试角度看，你必须能区分调查、观察性研究和设计实验。调查在不对变量进行操控的情况下从样本中收集信息，而实验则有意识地施加处理以测量其对响应变量的影响。方法的选择直接影响因果结论的强度以及混杂因素的有无。
The CIE syllabus highlights that you should be able to plan an investigation: formulate a hypothesis, identify the population of interest, decide on appropriate sampling techniques, and consider practical and ethical constraints. In practical assessment items, you may be asked to critique a given investigation plan and suggest improvements to reduce bias and increase reliability.
CIE 大纲强调你应能规划一项调查：提出假设、确定目标总体、选择合适的抽样方法，并考虑实际与伦理约束。在实践类考题中，你可能需要评价给定的调查方案并建议改进，以减少偏差、提高可靠性。
2. Types of Data and Their Implications | 数据类型及其影响
Data can be classified as categorical (nominal or ordinal) or numerical (discrete or continuous). Recognising the type of data is crucial because it dictates which statistical diagrams and measures are appropriate. For instance, the mode is the only meaningful average for nominal data, while the median can be used for ordinal data, and all three averages are suitable for symmetric numerical data.
数据可分为分类数据（名义或有序）和数值数据（离散或连续）。识别数据类型至关重要，因为它决定了哪些统计图表和度量是合适的。例如，对名义数据只有众数有意义，对有序数据可用中位数，而对对称数值数据三种平均数都可使用。
In an experimental setting, the response variable is almost always numerical, but explanatory variables can be categorical (e.g., treatment group vs control group) or numerical (e.g., dosage). Exam questions often ask you to identify the variable type and to justify your choice of measure, so be precise in using terms like ‘qualitative’ and ‘quantitative’ only when the context requires them.
在实验环境中，响应变量几乎总是数值型，但解释变量可以是分类的（如处理组与对照组）或数值的（如剂量）。考试常会要求你识别变量类型并说明选择度量的理由，因此要注意在上下文中恰当使用“定性”与“定量”等术语。
3. Sampling Methods: Randomness Matters | 抽样方法：随机性是关键
Simple random sampling gives every member of the population an equal chance of selection, and every sample of the same size is equally likely. It requires a sampling frame – a complete list of all population members – and a method to select units without bias, such as using random number tables or computer-generated numbers. In practice, examiners want you to explain how randomness removes selection bias.
简单随机抽样使总体中每个成员被选中的概率相等，且每个相同大小的样本等可能被抽到。它需要一个抽样框——所有总体成员的完整名单——以及一个无偏的选择方法，如使用随机数表或计算机生成的数字。实践中，考官希望你能解释随机性如何消除选择偏差。
Stratified sampling divides the population into distinct strata based on a characteristic known to influence the response (e.g., age, gender, income). Within each stratum, a random sample is taken. This method ensures representation of all subgroups and can reduce sampling error compared with simple random sampling. In an exam, you might be asked to calculate the required stratum size using proportional allocation: n_i = n × (N_i / N).
分层抽样根据已知影响响应的特征（如年龄、性别、收入）将总体划分为互不相交的层，然后在每层内随机抽样。该方法保证了所有子群体的代表性，与简单随机抽样相比可减小抽样误差。考试中可能要求你用比例分配计算所需层样本量：n_i = n × (N_i / N)。
Systematic sampling selects every kth unit from a list after a random start, while cluster sampling treats natural groups (clusters) as primary sampling units, often sampling all members within selected clusters. Quota sampling, commonly used in opinion polls, is non‑probability: interviewers fill predetermined quotas of rather haphazardly chosen respondents. Be ready to discuss the advantages and disadvantages of each, with particular focus on cost, convenience, and potential bias.
系统抽样从名单中随机起点后每隔 k 个单元选取样本，整群抽样则将自然群体（群）作为初级抽样单元，常对选中群内所有成员进行调查。配额抽样常用于民意调查，属于非概率抽样：访员根据预定配额随意选取受访者。准备讨论每种方法的优缺点，重点关注成本、便利性及潜在偏差。
4. Questionnaire and Question Design | 问卷与问题设计
Because CIE papers sometimes present a scenario where a questionnaire is used to collect data, you need to recognise good practice in question design. Questions should be clear, concise, unambiguous, and free from leading or emotionally charged language. Avoid double‑barrelled questions that ask about two issues in one. Response options should be exhaustive and mutually exclusive, especially for closed‑ended questions.
由于 CIE 试卷有时会给出通过问卷收集数据的情境，你需要识别问题设计的良好做法。问题应当清晰、简洁、无歧义，并避免引导性或带有情绪色彩的语言。避免一个问句询问两个问题的“双管”问题。回答选项应穷尽且互斥，尤其是封闭式问题。
Piloting a questionnaire on a small group helps identify confusing questions and ensures that the data collected will be valid and reliable. The order of questions matters too: start with straightforward, non‑sensitive items to build trust. Exam questions frequently ask you to suggest improvements to a poorly designed questionnaire, so learn how to spot vague wording, overlapping categories, and missing time frames.
对问卷进行小范围预试有助于发现令人困惑的问题，并确保收集的数据有效且可靠。问题的顺序也很重要：从简单、不敏感的项目开始以建立信任。考试常会要求你对一份设计不当的问卷提出改进建议，因此要学会识别模糊措辞、重叠的类别以及缺少时间范围等问题。
5. Core Principles of Experimental Design | 实验设计的核心原则
The three fundamental principles of designed experiments are randomisation, replication, and control. Randomisation assigns experimental units to treatment groups by chance, which averages out the effects of lurking variables and prevents systematic bias. Without randomisation, any observed difference might be due to confounding rather than the treatment itself.
设计实验的三项基本原则是随机化、重复和对照。随机化通过机会将实验单元分配到处理组，以平均掉潜在变量的影响并防止系统偏差。没有随机化，任何观察到的差异可能源自混杂，而非处理效应本身。
Replication means applying each treatment to several independent units, not simply taking multiple measurements on one unit. It provides an estimate of experimental error and increases the precision of treatment effect estimates. The number of replicates must be chosen with practical constraints in mind; more replication reduces the impact of anomalous results but may be costly or time‑consuming.
重复意味着将每种处理应用于多个独立单元，而不仅仅是同一个单元的多次测量。它提供了实验误差的估计，并提高了处理效应估计的精度。重复数必须在实际约束下选择；更多的重复可以减少异常结果的影响，但可能成本高、耗时长。
Control in experiments takes two forms: inclusion of a control group that receives no treatment or a standard treatment, which provides a baseline for comparison; and controlling extraneous variables by holding them constant wherever possible. For example, in a crop yield experiment, soil type, watering schedule, and sunlight exposure should be kept the same for all plots except for the fertiliser treatment.
实验中的对照有两种形式：纳入一个不接受处理或接受标准处理的对照组，它为比较提供基线；以及尽可能保持外来变量不变，以此控制它们。例如，在作物产量实验中，土壤类型、浇水计划和日照时间除了施肥处理外，对所有地块应保持一致。
6. Blocking and Paired Comparisons | 区组与配对比较
Blocking is used to account for a known source of variation that cannot be eliminated. Units are grouped into blocks (e.g., age groups, field locations) so that within‑block variability is small, and treatments are then randomly assigned inside each block. This design reduces the residual error and makes treatment comparisons more sensitive. The paired design is a special case: each pair is a block of size two, with one unit receiving treatment A and the other treatment B.
区组设计用于应对无法消除的已知变异来源。将单元分入区组（如年龄组、田间位置），使组内变异较小，然后在每个区组内随机分配处理。这种设计减少了残差误差，使处理比较更灵敏。配对设计是其特例：每个对子即大小为二的区组，一个单元接受处理 A，另一个接受处理 B。
When a question presents a matched‑pairs scenario – such as before‑and‑after measurements on the same subject – you should recognise that the analysis will focus on the differences within each pair. This eliminates the between‑subject variability that does not relate to the treatment. CIE exam questions often test your ability to suggest a blocking factor that would improve precision, so be prepared to identify variables like gender, age, or baseline health that could influence the response.
当题目给出配对情境——如同一个体的前后测量——你应认识到分析将关注每对内部的差值。这消除了与处理无关的个体间变异。CIE 考题经常测试你能否提出一个可提高精度的区组因素，因此要准备好识别性别、年龄或基线健康状况等可能影响响应的变量。
7. Bias, Confounding, and the Placebo Effect | 偏差、混杂与安慰剂效应
Bias is systematic error that causes results to deviate from the truth. Selection bias occurs when the sample is not representative; measurement bias arises from faulty instruments or leading questions; non‑response bias happens when those who refuse to participate differ systematically from those who agree. The examiner expects you to name the specific type of bias and explain how it could distort conclusions.
偏差是导致结果偏离真实值的系统误差。选择偏差发生在样本不具有代表性时；测量偏差源自仪器故障或诱导性问题；无应答偏差则发生在拒绝参与者与同意参与者存在系统差异时。考官期望你能说出具体的偏差类型，并解释它将如何扭曲结论。
Confounding is a particularly dangerous threat in observational studies and poorly designed experiments. A confounding variable is one that is associated with both the explanatory variable and the response, making it impossible to separate their effects. For example, if heavy coffee drinkers also tend to smoke more, any observed link between coffee and heart disease might be confounded by smoking. Randomisation is the only sure way to break the link between treatment assignment and potential confounders.
混杂在观察性研究和设计糟糕的实验中尤为危险。混杂变量是指与解释变量和响应变量都相关的变量，导致无法区分它们各自的影响。例如，如果大量饮用咖啡者往往也吸烟更多，那么所观察到的咖啡与心脏病之间的关联可能被吸烟所混杂。随机化是打破处理分配与潜在混杂因子之间联系的唯一可靠方法。
The placebo effect describes a psychological or physiological response to a dummy treatment, driven by the patient’s belief in its efficacy. In medical trials, blinding – keeping subjects unaware of which treatment they receive – and double‑blinding, where both subjects and the assessors are masked, help to mitigate this effect. Questions often ask you to justify why a placebo control is necessary: it allows the true treatment effect to be measured net of any psychological influences.
安慰剂效应描述的是患者对虚假治疗产生的心理或生理反应，由患者对疗效的信念驱动。在医学试验中，盲法——使受试者不知其接受的是何种处理——以及双盲法（受试者与评估者均不知情）有助于减轻这种效应。题目常要求你论证为什么安慰剂对照是必要的：它可以排除任何心理影响，测量到净处理效应。
8. Comparative and Factorial Structures | 比较与析因结构
Almost all good experiments are comparative: they compare two or more treatments, not just observe a single group. A completely randomised design compares several treatments by allocating units purely at random. When two factors are of interest, a factorial design crosses all levels of factor A with all levels of factor B, enabling the investigation of interactions – situations where the effect of one factor depends on the level of the other.
几乎所有好的实验都是比较性的：它们比较两种或更多处理，而不仅仅是观察一组。完全随机化设计通过纯粹随机分配来比较若干处理。当关注两个因子时，析因设计将因子 A 的所有水平与因子 B 的所有水平交叉，从而能够研究交互作用——即一个因子的效应依赖于另一个因子的水平的情况。
Even at AS Level, you might be asked to recognise the structure of a simple factorial experiment and to comment on how it improves efficiency. Testing factors simultaneously can be more economical than running separate experiments, and it also reveals interactions that single‑factor experiments would miss. Understand the notation: a 2×3 factorial has two levels of the first factor and three of the second, yielding six treatment combinations.
即使在 AS 阶段，你仍可能被要求识别简单析因实验的结构并评论它如何提高效率。同时测试多个因子比单独进行实验更经济，并且还能揭示单因子实验会遗漏的交互作用。要理解记号：一个 2×3 析因设计表示第一个因子有 2 个水平，第二个有 3 个水平，共产生 6 个处理组合。
9. Ethical and Practical Constraints | 伦理与实际约束
Real‑life investigations are subject to ethical guidelines and practical limitations, and CIE examiners value your awareness of these. Ethical issues include informed consent, privacy and confidentiality of data, the right to withdraw, and avoidance of harm. In human experiments, it must be ethically acceptable to randomly assign treatments; if not, an observational study is more appropriate.
现实调查受伦理准则和实际限制的约束，CIE 考官重视你对这些方面的认知。伦理问题包括知情同意、数据隐私与保密、退出权以及避免伤害。在人体实验中，随机分配处理必须在伦理上可接受；否则观察性研究更为合适。
Practical constraints include cost, time, availability of subjects, and the feasibility of measuring the response accurately. When answering a question that asks ‘What factors should the researcher consider when designing this experiment?’, be sure to mention both statistical principles and real‑world constraints. For example, using a larger sample is desirable but may be limited by budget; blocking on a rare characteristic may be ideal but impossible to implement if the blocks cannot be identified in advance.
实际约束包括成本、时间、被试的可获得性以及准确测量响应的可行性。在回答“研究人员在设计该实验时应考虑哪些因素？”这类问题时，一定要同时提及统计原则与现实约束。例如，更大的样本量是理想的，但可能受预算限制；按罕见特征区组可能是最好的，但如果无法提前识别区组便不可行。
10. Data Collection and Recording | 数据收集与记录
Accurate and systematic data recording is the bedrock of reproducible research. In practical assessments, you may be presented with raw data sheets, tally charts, or database extracts and asked to identify errors such as impossible values, missing entries, or inconsistent coding. Always check data ranges and logical relationships before performing any analysis.
准确、系统的数据记录是可重复研究的基石。在实践评价中，你可能会看到原始数据表、计数表格或数据库摘录，并被要求识别错误，例如不可能的值、缺失条目或不一致的编码。在进行任何分析之前，务必检查数据范围和逻辑关系。
When designing a data collection form, include columns for date, time, observer or experimental unit ID, measured variables, and any relevant conditions (e.g., temperature, humidity). Clear labels and units avoid ambiguity. In exam settings, you might be asked to design a brief data collection sheet; be explicit about units and permissible values, and leave space for comments that could explain unexpected results.
在设计数据收集表时，要纳入日期、时间、观察者或实验单元编号、测量变量以及任何相关条件（如温度、湿度）等列。清晰的标签和单位可避免歧义。在考试情境中，你可能被要求设计一张简单的数据收集表；要明确单位和允许值，并留出备注空间，以便解释意外结果。
11. Common Mistakes and How to Avoid Them | 常见错误及其避免方法
One frequent pitfall is confusing the experimental unit with the observation unit. If a treatment is applied to a whole class but data are recorded on individual students within the class, the class is the experimental unit. Falsely treating individual students as independent replicates inflates the apparent sample size and produces misleadingly small standard errors – a problem known as pseudoreplication.
一个常见误区是混淆实验单元与观测单元。如果处理施加于整个班级，但数据则记录在班级内的每个学生身上，那么班级才是实验单元。错误地将单个学生视为独立重复会增大表面样本量，并产生误导性地极小的标准误——这一问题被称为伪重复。
Another common error is neglecting to blind assessors in subjective measurements. If the person measuring the response knows which treatment was applied, expectation bias can creep in. Whenever possible, use objective instruments and standardised protocols to reduce measurement bias. In the exam, always ask yourself: ‘Who could influence the result, and how can that influence be removed?’
另一个常见错误是在主观测量中忽略对评估者的盲法。如果测量响应的人知道施加的是哪种处理，期望偏差便会潜入。只要可能，应使用客观仪器和标准化方案以减少测量偏差。考试中，始终问自己：“谁会影响到结果？这种影响如何消除？”
12. Exam Technique: Structuring Your Answer | 考试技巧：答案的结构
When faced with an open‑ended question about designing or critiquing an investigation, use a checklist approach in your mind: population and sampling method, treatment structure, randomisation, replication, control of extraneous variables, blinding, ethical considerations, and data analysis plan. Avoid listing bullet points superficially; instead, explain the rationale for each choice and link it to the specific context in the question.
当遇到设计或评价调查的开放性问题时，可在脑中用清单法：总体与抽样方法、处理结构、随机化、重复、外来变量的控制、盲法、伦理考虑以及数据分析计划。避免肤浅地罗列要点；相反，要解释每一项选择的理由，并将其与题目中的具体情境联系起来。
Full‑mark answers typically do two things well: they give statistical terminology correctly and they apply it to the scenario. For example, instead of saying ‘use a large sample’, write ‘increase the number of replicates to reduce the standard error of the estimated treatment effect and to improve the power of detecting a meaningful difference’. Practice translating everyday language into precise statistical language, and you will stand out in the CIE marking scheme.
高分答案通常做好两件事：准确使用统计术语，并将其应用于情境。例如，与其说“使用大样本”，不如写“增加重复数以减小估计处理效应的标准误，并提高检测有意义差异的效力”。练习将日常语言转化为精确的统计语言，你就能在 CIE 评分方案中脱颖而出。
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Cancel reply