AS Cambridge Statistics: Key Points for Experimental/Practical Assessment | AS 剑桥统计：实验/实践考核要点

📚 AS Cambridge Statistics: Key Points for Experimental/Practical Assessment | AS 剑桥统计：实验/实践考核要点

Mastering the practical component in AS Cambridge Statistics means moving beyond formulae to real-world investigation. Whether you are designing a simple experiment to compare reaction times or carrying out a questionnaire survey, the examiner is looking for clear hypothesis statements, thorough planning, controlled data collection, and insightful evaluation. This guide breaks down the key assessment points to help you structure your project, avoid common pitfalls, and present your findings in a way that demonstrates genuine statistical thinking.

要在 AS 剑桥统计的实践考核中取得好成绩，光会套公式远远不够。无论是设计一个比较反应时间的简单实验，还是开展问卷调查研究，考官关注的是清晰的假设陈述、周密的计划、受控的数据采集以及深入的评价。这份指南将逐一拆解核心考核要点，帮助你搭建项目框架、避开常见陷阱，并以展现真实统计思维的方式呈现你的发现。

---

1. Understanding the Practical Assessment Structure | 理解实践考核的结构

AS Statistics practical assessments typically require you to plan and carry out a small-scale investigation. The task may be presented as a problem statement or a scenario, leaving you to decide how to collect and analyse data. Marks are awarded for planning, implementation, data presentation, analysis, interpretation, and evaluation, with close attention paid to the logic behind every choice you make.

AS 统计的实践考核通常要求你计划并完成一项小规模调查。题目可能以问题陈述或情境的形式给出，由你决定如何采集和分析数据。分数分布在计划、实施、数据展示、分析、解释和评价这几个环节，考官会格外关注你每一步选择背后的逻辑。

• Internal assessment (school-based): Often conducted as a task under controlled conditions, but preparatory work may be done in advance. / 校内评估：通常为在受控条件下完成的任务，但预备工作可以提前进行。
• Exam paper style (Paper 6): Some Cambridge specifications include a written paper that tests experimental design and data interpretation through scenario-based questions. / 笔试形式（如 Paper 6）：某些剑桥大纲设有专门考查实验设计和数据解读的笔试，以情景题形式出现。
• Key criteria: Clarity of aim, identification of variables, choice of sampling or experimental technique, data handling accuracy, and critical evaluation. / 关键评分标准：目标清晰度、变量的识别、抽样或实验技术的选择、数据处理的准确性以及批判性评价。

---

2. Formulating Research Questions and Hypotheses | 提出研究问题与假设

A sharp, focused research question is the engine of any statistical investigation. Start with a general curiosity, then narrow it down to something measurable. For example, ‘Do girls and boys have different reaction times?’ can become ‘Is there a difference in the mean reaction time of 16‑year‑old female and male students?’ Alongside this, state a null hypothesis (H₀) and an alternative hypothesis (H₁). The null typically asserts no difference or no correlation, while the alternative represents the effect you suspect.

一个明确、聚焦的研究问题是任何统计调查的引擎。从宽泛的好奇心出发，然后将其缩小到可量化的范围。例如，“女生和男生的反应速度不同吗？”可以转化为“16 岁男女学生的平均反应时间是否存在差异？”与此同时，明确陈述零假设 (H₀) 和备择假设 (H₁)。零假设通常声称没有差异或没有相关性，备择假设则代表你猜测可能存在的效应。

• Use clear, operational definitions: ‘reaction time’ measured in seconds using an online test. / 使用清晰、可操作的定义：“反应时间”用在线测试以秒为单位测量。
• Always include a justification for your hypothesis based on prior knowledge or a logical argument. / 始终基于先验知识或逻辑论证为你的假设提供一个理由。
• Avoid vague or untestable hypotheses: ‘people like fast food’ cannot be objectively measured without clear criteria. / 避免模糊或不可检验的假设：“人们喜欢快餐”若无明确标准便无法客观测量。

---

3. Identifying Independent, Dependent, and Control Variables | 识别自变量、因变量与控制变量

In any experiment, clearly distinguish the variable you deliberately change (independent variable), the variable you measure to see the effect (dependent variable), and all other factors that must be kept constant to ensure a fair test (control variables). Confusing these leads to flawed results and loss of marks.

在任何实验中，清晰区分你刻意改变的变量（自变量）、你为观察效果而测量的变量（因变量）以及所有必须保持恒定以确保公平测试的其他因素（控制变量）。混淆这些概念会导致结果失真并丢分。

Variable Type | 变量类型	Example (Reaction time & Caffeine) | 示例（反应时间与咖啡因）
Independent | 自变量	Amount of caffeine consumed (e.g. none vs. one 330ml can of caffeinated drink) | 摄入的咖啡因量（如无咖啡因 vs. 一罐 330ml 含咖啡因饮料）
Dependent | 因变量	Mean reaction time (seconds) on a ruler drop test | 直尺跌落测试的平均反应时间（秒）
Control | 控制变量	Same test environment, same ruler, same time of day, participants with similar sleep and food intake, no other stimulants | 相同测试环境、相同直尺、相同时间段、参与者睡眠与饮食相似、无其他兴奋剂

Even in questionnaires, distinguish explanatory and response variables. If you investigate whether hours of homework are associated with test scores, homework hours are the explanatory variable and test score is the response. Control for confounding variables like prior attainment by restricting the sample or recording the data.

即便在问卷研究中，也要区分解释变量与响应变量。如果你调查家庭作业时间是否与考试成绩相关，作业时间就是解释变量，考试成绩是响应变量。通过限制样本或记录数据来控制诸如先前学业水平等混杂变量。

---

4. Designing Experiments: Randomisation and Replication | 实验设计：随机化与重复

A well‑designed experiment includes random allocation of subjects to treatment groups, where possible. Randomisation helps balance out unknown confounding variables. Replication means using a sufficient number of participants or trials, not just measuring one person once. This allows you to estimate variability and increases the reliability of your conclusions.

一个设计良好的实验在可能的情况下应包含将受试者随机分配到处理组。随机化有助于平衡未知的混杂变量。重复意味着使用足够数量的参与者或试验次数，而不是只测量一个人一次。这使你能够估计变异性，并提高结论的可靠性。

• Completely randomised design: All subjects randomly assigned to treatment or control group. / 完全随机设计：所有受试者随机分配至处理组或对照组。
• Paired or matched design: Pairs of similar subjects (e.g. twins, pre‑test/post‑test) receive different treatments, reducing variability. / 配对或被试内设计：相似的受试者对（如双胞胎、前测/后测）接受不同处理，减少变异性。
• Blocking: Divide subjects into homogeneous blocks (e.g. by age group) and randomise within blocks to control known nuisance factors. / 区组化：将受试者分为同质的区组（例如按年龄段），在区组内随机化，以控制已知的干扰因素。
• Always describe your randomisation method precisely: ‘I will assign each volunteer a number and then use a random number table to allocate them to group A or B.’ / 始终精确描述你的随机化方法：“我会给每位志愿者分配一个编号，然后使用随机数字表将他们分配到 A 组或 B 组。”

---

5. Sampling Methods: Getting a Representative Picture | 抽样方法：获取代表性样本

If your investigation involves a survey, the way you choose participants determines how far you can generalise results. Understand the strengths and limitations of different sampling techniques. For an AS practical, you might use convenience or quota sampling for feasibility, but you must acknowledge the associated biases.

如果你的调查涉及问卷，你选择参与者的方式决定了结果有多大的推广度。要理解不同抽样技术的优缺点。对于 AS 实践，你可能会因可行性而使用便利抽样或配额抽样，但你必须承认由此带来的偏差。

Method | 方法	Description | 描述	Bias potential | 潜在偏差
Simple random | 简单随机	Every member of population has equal chance of selection. | 总体每个成员有同等机会被选中。	Low, but may be impractical without sampling frame. | 低，但没有抽样框可能不实际。
Stratified | 分层	Population split into strata, random sample taken from each. | 总体分成层，每层随机抽取样本。	Very low if strata well defined; ensures representation. | 如果层定义良好则非常低，确保代表性。
Systematic | 系统	Select every kth individual from a list. | 从列表中每隔 k 个选取一个个体。	Possibly if list has hidden periodicity. | 若列表存在隐藏周期性可能出现。
Convenience | 便利	Select whoever is easily available. | 选择最容易接触到的人。	High; sample may not reflect population. | 高，样本可能无法反映总体。

State clearly your target population (e.g. all Year 12 students in the school) and how you will obtain your sample. Acknowledge any mismatch between the sampling frame and target population.

清晰陈述你的目标总体（例如全校所有 12 年级学生）以及你将如何获取样本。承认抽样框与目标总体之间的任何不匹配。

---

6. Data Collection Tools: Questionnaires, Observations, and Measurement | 数据收集工具：问卷、观察与测量

Choose tools that minimise measurement error and maximise relevance. If using a questionnaire, pilot it with a small group to catch ambiguous wording. Questions should be clear, brief, and designed without leading the respondent. For observations and measurements, define exactly how and when you will record data, and use calibrated instruments if possible.

选择能最小化测量误差、最大化相关性的工具。如果使用问卷，先在小群体中试测，以发现模棱两可的措辞。问题应清晰、简短，且设计时不具有诱导性。对于观察和测量，确切定义你将如何、何时记录数据，并尽可能使用经过校准的仪器。

• Closed questions produce numerical or categorical data that can be easily summarised; open questions give richer insight but are harder to analyse. / 封闭式问题产生易于总结的数值或分类数据；开放式问题能提供更丰富的见解但分析困难。
• Avoid double-barrelled questions: ‘Do you enjoy maths and find the lessons useful?’ should be split into two separate items. / 避免双重问题：“你喜欢数学并觉得课程有用吗？”应拆分为两个独立条目。
• In experiments, take repeated measurements to reduce random errors: three trials for reaction time per condition, then average. / 在实验中，多次测量以减少随机误差：每种条件下进行三次反应时间测试，然后取平均值。
• Record data immediately in a well-structured table or spreadsheet to prevent transcription errors. / 立即将数据记录在设计良好的表格或电子表格中，以防止誊写错误。

---

7. Organising and Presenting Data Neatly | 整洁地整理与展示数据

Raw data is rarely informative in its original form. Tidy tables with clear headings and consistent units are the first step. For larger datasets, use a stem-and-leaf diagram to show the shape while preserving original values, or a frequency table with class intervals. Graphs should be selected according to data type: bar charts for categorical data, histograms for continuous, and scatter diagrams for bivariate relationships.

原始数据在其原本形态下很少能提供信息。带有清晰表头和一致单位的整洁表格是第一步。对于较大的数据集，使用茎叶图展示分布形状同时保留原始数值，或使用带有组距的频数表。图表应根据数据类型选择：条形图用于分类数据、直方图用于连续数据、散点图用于双变量关系。

• Always label axes fully (‘Number of hours of sleep’ not just ‘hours’), include units, and give the graph a descriptive title. / 始终完整标记坐标轴（“睡眠小时数”而不只是“小时”），包含单位，并为图表提供一个描述性标题。
• Histograms need the vertical axis to be frequency density when class widths are unequal; otherwise frequency is acceptable. / 当组距不等时，直方图的纵轴必须是频数密度；否则频数也可接受。
• For comparisons, side‑by‑side box plots are excellent for highlighting medians, spreads, and outliers. / 为了进行比较，并列箱形图是突出中位数、散布和异常值的绝佳选择。
• Avoid 3D effects or unnecessary decoration on graphs – simplicity wins. / 避免图表中的 3D 效果或不必要的装饰——简洁致胜。

---

8. Descriptive Statistics: Summarising Central Tendency and Spread | 描述统计：概括集中趋势与离散程度

Use appropriate summary statistics: mean and standard deviation for symmetric data without outliers, median and interquartile range (IQR) for skewed data or when outliers are present. Show your working for any calculated values, but also feel free to use calculator functions accurately – the key is to interpret, not just compute.

使用恰当的汇总统计量：对于无异常值的对称数据使用均值和标准差，对于偏态数据或存在异常值时使用中位数和四分位距 (IQR)。对于任何计算量，展示你的计算步骤，但也可放心使用计算器功能——关键在于解读，而不仅仅是计算。

Mean x̄ = Σx / n    s = √[ Σ(x − x̄)² / (n − 1) ]    IQR = Q₃ − Q₁

• Identify and discuss any outliers using the 1.5 × IQR rule or two standard deviations from the mean. / 使用 1.5 × IQR 法则或距均值两个标准差的方法识别并讨论任何异常值。
• Comment on what the measures reveal: a large standard deviation indicates high variability in results. / 评论这些度量揭示了什么：较大的标准差表明结果变异性高。
• For categorical data, give proportions or percentages in each category, and use the mode where appropriate. / 对于分类数据，给出每个类别的比例或百分比，并酌情使用众数。

---

9. Drawing Inferences and Conclusions Tied to the Hypothesis | 得出与假设相关的推断与结论

Your conclusion must directly address the hypothesis. Do not claim to ‘prove’ anything; instead, discuss whether the evidence supports H₁ or not. Use appropriate statistical language: for example, if you compare two means, you might perform a two-sample t‑test (or discuss its logic even at AS level where formal hypothesis testing may be introduced gently). At minimum, compare summary statistics and acknowledge overlap in spreads.

你的结论必须直接回应假设。不要宣称“证明”了什么；相反，讨论证据是否支持 H₁。使用恰当的统计语言：例如，如果你比较两个均值，你可能会进行双样本 t 检验（或即便在 AS 阶段仅温和引入正式假设检验，也要讨论其逻辑）。至少，比较汇总统计量并承认分布的分散程度存在重叠。

• If a t‑test is used, state the p‑value and whether it is below the chosen significance level (e.g. p < 0.05). / 如果使用 t 检验，说明 p 值以及它是否低于选定的显著性水平（如 p < 0.05）。
• No significant difference does not mean the null hypothesis is true; it means there is insufficient evidence to reject it. / 没有显著差异并不意味着零假设为真，它意味着没有足够证据拒绝零假设。
• Relate findings to real-world context: ‘The observed difference of 0.03 seconds may be small in practical terms, even if statistically significant.’ / 将发现与实际情境联系起来：“即使统计上显著，观察到的 0.03 秒差异在实际意义上可能很小。”

---

10. Evaluation: Limitations, Reliability, and Validity | 评价：局限性、信度与效度

A top‑band evaluation goes beyond listing ‘I could have used more participants’. It links specific limitations to the validity or reliability of the results and suggests realistic improvements. Discuss both systematic and random errors, and reflect on whether the design genuinely tested what it set out to test.

高水平的评价远不止罗列“我本可以增加参与者”。它将具体的局限性同结果的信度或效度联系起来，并提出切实可行的改进建议。讨论系统误差和随机误差，反思这个设计是否真正测试了它想要测试的东西。

• Internal validity: Were the control variables genuinely held constant? Could there be a confounding variable? / 内部效度：控制变量是否真的保持恒定？会不会存在混杂变量？
• External validity: How far can you generalise? A convenience sample of friends limits generalisability. / 外部效度：你能推广到多大程度？便利样本的朋友限制了普遍性。
• Measurement reliability: If you repeated the experiment tomorrow, would you get similar results? Equipment precision and human reaction might introduce inconsistency. / 测量信度：如果你明天重复这个实验，会得到相似结果吗？设备精度和人的反应可能引入不一致性。
• Suggest specific improvements: ‘Instead of a manual stopwatch, use an electronic sensor connected to a computer to record reaction times to the nearest 0.001 seconds.’ / 提出具体改进：“代替手动秒表，使用连接电脑的电子传感器，以精确到 0.001 秒记录反应时间。”

---

11. Writing a Coherent Practical Report | 撰写条理清晰的实践报告

Structure your report so that the reader can follow your logical journey: Introduction (aim, hypothesis, background), Method (design, sampling, materials), Results (tables, graphs, summary statistics), Analysis (interpretation, comparison), Conclusion, and Evaluation. Use clear headings and write in the third person passive where possible to maintain a formal tone.

组织你的报告，让读者能跟随你的逻辑旅程：引言（目的、假设、背景）、方法（设计、抽样、材料）、结果（表格、图表、汇总统计量）、分析（解读、比较）、结论以及评价。使用清晰的小标题，并尽可能使用第三人称被动语态以保持正式语气。

• Avoid personal anecdotes: ‘I felt nervous’ is irrelevant; instead, ‘Potential variability may arise from participant fatigue.’ / 避免个人轶事：“我感到紧张”无关紧要；而应说“潜在的变异性可能来自参与者疲劳。”
• Number all tables and figures, and refer to them in the text: ‘Figure 1 shows a histogram of the response times for both groups.’ / 给所有表格和图表编号，并在文中引用它们：“图 1 显示了两个组的反应时间直方图。”
• Check that raw data, worked calculations, and a blank copy of the questionnaire appear in appendices if required. / 若要求，检查原始数据、计算过程以及空白问卷是否出现在附录中。
• Proofread for units consistency: saying ‘cm’ in the text but recording ‘mm’ in tables creates confusion. / 校对单位一致性：文中说“厘米”但表格记录“毫米”会造成混乱。

---

12. Ethical Considerations When Working with Human Participants | 涉及人类参与者的伦理考虑

Ethical practice is not just a box‑checking exercise; it strengthens the trustworthiness of your study. Always obtain informed consent, explain the purpose of the investigation in plain language, and guarantee anonymity for survey responses. Allow participants to withdraw at any time without penalty, and debrief them afterwards, especially if any deception was used (which is rarely necessary at AS level).

伦理操作不只是一个打勾的练习，它能增强你研究的可信度。始终获取知情同意，用通俗语言解释调查的目的，并保证问卷回复的匿名性。允许参与者在任何时候无条件退出，并在结束后向他们解释研究情况，特别是如果使用了任何形式的欺骗（在 AS 阶段这很少有必要）。

• For experiments involving mild stress or physical activity, risk‑assess and ensure no harm comes to participants. / 对于涉及轻微压力或身体活动的实验，进行风险评估并确保参与者不受伤害。
• Store data securely and do not share identifiable information. / 安全地存储数据，不分享可识别信息。
• If your investigation involves sensitive topics (e.g. body image, family income), be particularly careful in question design and debriefing. / 如果你的调查涉及敏感话题（如体型映像、家庭收入），在问题设计和事后解释中要格外小心。
• Mentioning ethical steps taken demonstrates maturity in your write‑up and is rewarded in evaluation sections. / 在报告中提及所采取的伦理步骤能展现你的成熟度，并在评价部分获得嘉许。

---

Published by TutorHao | Statistics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）