📚 A-Level Mathematics: Experimental Operations Guide | A-Level 数学：实验操作指南
In A-Level Mathematics, particularly within the Statistics components, experiments form the backbone of data-driven decision-making. Whether you are designing a clinical trial, testing a new fertiliser, or exploring the effect of light on plant growth, a solid grasp of experimental design principles is essential. This guide unpacks the key ideas behind conducting valid experiments, from randomisation and control to data collection and common pitfalls, helping you tackle both coursework and exam questions with confidence.
在A-Level数学中，尤其是在统计学模块里，实验是数据驱动决策的基础。无论你是在设计临床试验、测试新肥料的效果，还是研究光照对植物生长的影响，牢固掌握实验设计原则都至关重要。本文详细解读了开展有效实验的关键思路，涵盖随机化、对照、数据收集及常见易错点，帮助你在课程作业和考试中自信应对。
1. What is an Experiment in A-Level Maths? | A-Level数学中的实验是什么？
An experiment in the statistical sense is a planned investigation in which the researcher deliberately imposes one or more treatments on experimental units to observe and compare responses. Unlike an observational study where we merely record what happens, an experiment allows us to infer causation.
在统计意义上，实验是一种有计划的研究：研究者故意对实验单元施加一个或多个处理，以观察和比较响应结果。与仅仅记录现象的观察性研究不同，实验能够让我们推断因果关系。
For instance, to test whether a new teaching method improves exam scores, you would randomly assign students to either the new method or a traditional one, teach them, and then compare results. This active manipulation distinguishes an experiment from a survey.
例如，要检验一种新教学方法是否能提高考试成绩，你可以将学生随机分配到新方法组或传统方法组，开展教学，然后比较结果。这种主动的操控正是实验与调查的区别所在。
In A-Level Mathematics, you will often be asked to describe how to set up such an experiment, and to explain why certain features – like randomisation – are crucial for valid conclusions.
在A-Level数学中，你经常会被要求描述如何设计这样的实验，并解释为什么某些特征（如随机化）对于得出有效结论至关重要。
2. Key Principles of Experimental Design | 实验设计的关键原则
The three fundamental principles that underpin any reliable experiment are randomisation, replication, and control. These principles work together to reduce bias, account for variability, and allow for meaningful comparisons.
支撑任何可靠实验的三个基本原则是随机化、重复和对照。这些原则共同作用，以减少偏倚、考虑变异性，并支持有意义的比较。
Randomisation ensures that each experimental unit has an equal chance of receiving any treatment. This prevents systematic differences between groups and is the best defence against confounding variables.
随机化确保每个实验单元有均等的机会接受任意一种处理。这可以防止组间出现系统性差异，是抵御混杂变量的最佳手段。
Replication means applying each treatment to multiple experimental units. Without enough replicates, we cannot distinguish a genuine treatment effect from random noise.
重复是指将每种处理施加于多个实验单元。如果没有足够的重复，我们无法区分真正的处理效应和随机误差。
Control involves holding other variables constant or using a baseline (control group) for comparison. A control group may receive a placebo or standard treatment, providing a benchmark against which the new treatment is measured.
对照涉及保持其他变量恒定，或使用基线（对照组）进行比较。对照组可能接受安慰剂或标准处理，为衡量新处理的效果提供基准。
3. Randomisation and Its Importance | 随机化及其重要性
Randomisation is the cornerstone of statistical experimentation. It is not about haphazard choice but about using a chance mechanism – such as a random number table or a calculator’s random integer function – to allocate treatments.
随机化是统计实验的基石。它不是随意选择，而是利用随机机制（如随机数表或计算器的随机整数功能）来分配处理。
Why is this so critical? Without randomisation, groups might differ in ways we have not measured. For example, if you let volunteers choose between a new drug and a placebo, healthier people might opt for the drug, making it seem effective even if it is not. Random assignment breaks such links.
为什么这一点如此重要？如果不随机化，各组可能在某些我们未测量的方面存在差异。例如，如果让志愿者在新药和安慰剂之间自行选择，健康的人可能倾向选择新药，这样即使新药无效也会显得有效。随机分配能切断这类关联。
In your A-Level exam, you might be asked to explain how to randomly assign 30 plants to three fertiliser treatments. A simple method: number the plants 1 to 30, use a calculator to generate random numbers 1–30 without replacement, and assign the first ten numbers to treatment A, the next ten to B, and the remainder to C.
在A-Level考试中，你可能会被要求解释如何将30株植物随机分配到三种肥料处理中。一个简单的方法是：给植物编号1至30，用计算器生成1到30的不重复随机数，将前10个号码分配给处理A，随后10个给B，剩下的给C。
4. Control Groups and Blinding | 对照组与盲法
A control group provides the essential ‘what would have happened without the treatment’ scenario. In a clinical trial, the control group often receives a placebo – an inactive substance identical in appearance to the real drug.
对照组提供了关键的“如果没有处理会发生什么”的情景。在临床试验中，对照组通常接受安慰剂——一种外观与真药完全相同但无活性的物质。
Blinding further reduces bias. In a single-blind experiment, the subjects do not know which treatment they receive, preventing psychological effects. In a double-blind experiment, neither the subjects nor the experimenters know who gets which treatment until after the data are analysed, guarding against both subject and observer bias.
盲法进一步减少偏倚。在单盲实验中，受试者不知道自己接受的是哪种处理，从而避免心理效应。在双盲实验中，受试者和实验者都不知道谁接受了哪种处理，直到数据分析完成，这可以同时防范受试者偏倚和观察者偏倚。
When describing an experiment for A-Level, always mention whether blinding is feasible and why it matters. For instance, if you are testing the effect of music on concentration, it might be impossible to blind the participants, but you can still blind the person marking the concentration tests.
在为A-Level描述实验时，一定要提及是否可以进行盲法以及为什么重要。例如，如果你正在测试音乐对注意力的影响，可能无法让受试者盲试验，但你仍然可以对批改注意力测试的人实施盲法。
5. Replication and Sample Size | 重复与样本量
Replication – using several experimental units under each treatment – is what gives an experiment the power to detect real differences. A single plant, mouse, or student cannot represent an entire population.
重复——在每个处理下使用多个实验单元——使实验有能力检测出真正的差异。一株植物、一只老鼠或一个学生无法代表整个群体。
With more replicates, the estimate of the treatment effect becomes more precise. The standard error of the mean decreases, making confidence intervals narrower. Although exact sample size calculations may appear in Further Mathematics, at A-Level you should understand that larger samples generally yield more reliable results, provided they are randomly selected.
重复次数越多，对处理效应的估计就越精确。均值的标准误减小，置信区间变窄。虽然精确的样本量计算可能出现在进阶数学中，但在A-Level水平，你应该理解较大的样本通常能产生更可靠的结果，前提是它们是随机选取的。
A common exam question asks you to suggest an appropriate number of replicates and justify your choice. You might say, “Use 20 plants per treatment to balance practical constraints with the need to reduce random error.”
常见的考题要求你建议合适的重复次数并说明理由。你可以说：“每个处理使用20株植物，以在现实限制和减少随机误差之间取得平衡。”
6. Types of Experimental Designs | 实验设计的类型
Different situations call for different experimental designs. The table below summarises the most common ones encountered in A-Level Statistics.
不同情况下需要采用不同的实验设计。下表总结了A-Level统计学中最常见的几种设计。
| Design | 设计 | Description | 描述 | Advantages | 优点 |
|---|---|---|
| Completely Randomised Design (CRD) | 完全随机设计 | All experimental units randomly assigned to treatments. | Simple; suitable when units are homogeneous. |
| Randomised Block Design (RBD) | 随机区组设计 | Units grouped into blocks based on a known source of variability (e.g. age, soil type); within each block, random assignment to treatments. | Reduces variability due to blocking factor; increases precision. |
| Matched Pairs Design | 配对设计 | Subjects are paired up so that within each pair they are as similar as possible; one gets treatment A, the other B. | Controls for subject-to-subject variability; powerful with small samples. |
| Latin Square Design | 拉丁方设计 | Controls for two blocking factors simultaneously; treatments arranged in a square such that each appears once in each row and column. | Efficient when two nuisance factors exist; usually seen in more advanced contexts. |
In an A-Level problem, you might be asked to identify the design from a description or to suggest a design for a given scenario. For example, if a farmer wants to test three wheat varieties but his field has a fertility gradient, a randomised block design with blocks perpendicular to the gradient would be wise.
在A-Level题目中，你可能会被要求根据描述识别设计类型，或为特定情景建议一种设计。例如，如果一位农民想要测试三种小麦品种，但他的田地存在肥力梯度，那么采用区组方向与梯度垂直的随机区组设计是明智之举。
7. Collecting and Organising Data | 数据的收集与整理
Even the best experimental design fails if data collection is sloppy. Plan a data recording sheet before you start, listing the treatment, block (if any), and response variable for each unit.
如果数据收集草率，即使是最好的实验设计也会失败。在开始前设计好数据记录表，列出每个单元的处理、区组（如果有）和响应变量。
Use clear labels and, where possible, record measurements in consistent units. If multiple observers are involved, standardise the measurement procedure. For instance, if measuring plant height, agree whether to measure from the soil surface or the base of the stem.
使用清晰的标签，并尽可能用一致的单位记录测量值。如果有多名观察者参与，要统一测量程序。例如，测量株高时，要商定是从土表还是从茎基部量起。
Digital tools can help. A simple spreadsheet can store data, and many graphical calculators allow you to enter lists and immediately compute summary statistics. Always keep a backup of raw data; never replace original figures with calculated ones.
数字工具可以提供帮助。一个简单的电子表格就能存储数据，许多图形计算器允许输入列表并立即计算概要统计量。务必备份原始数据；绝不要用计算值替代原始数据。
8. Using Technology for Experiments | 运用技术进行实验
Technology plays a dual role in A-Level experiments: it helps with both design and analysis. Graphing calculators and software like GeoGebra can generate random numbers for randomisation, simulate probability distributions, and perform hypothesis tests.
技术在A-Level实验中扮演双重角色：它既能辅助设计，也能辅助分析。图形计算器和GeoGebra等软件可以生成用于随机化的随机数、模拟概率分布，并进行假设检验。
For example, to test whether a coin is fair, you can simulate 100 tosses on your calculator using the randBin function, record the number of heads, and repeat to build an empirical sampling distribution. This hands-on approach deepens understanding of p-values and significance.
例如，要测试一枚硬币是否公平，你可以在计算器上用randBin函数模拟100次抛掷，记录正面次数，并多次重复以构建经验抽样分布。这种动手操作的方法能加深对p值和显著性的理解。
When writing up your experiment, include screenshots or code snippets if allowed, but always explain what the technology is doing in statistical terms. For instance, “The calculator generated a random sample from a normal distribution with mean μ=50 and standard deviation σ=5.”
在撰写实验报告时，如果允许，可附上截图或代码片段，但一定要用统计术语解释技术在做什么。例如：“计算器从均值为μ=50、标准差为σ=5的正态分布中生成了一个随机样本。”
9. Common Mistakes in Experimental Work | 实验中的常见错误
A-Level examiners frequently report the same errors year after year. Understanding these pitfalls will sharpen your answers.
A-Level考官每年都会报告相同的错误。了解这些陷阱能让你的答案更加精准。
Confusing an experiment with an observational study: If you merely compare the exam scores of students who chose to attend revision classes with those who did not, that is not an experiment – it is an observational study, and you cannot claim causation.
将实验与观察性研究混淆：如果你仅仅是比较选择上复习课和没有上复习课的学生的考试成绩，那不是实验——那是一项观察性研究，你不能声称因果关系。
Ignoring confounding variables: Suppose you test a new energy drink by giving it to athletes in the morning and a placebo in the afternoon. Time of day is a confound; you cannot tell if any difference is due to the drink or the time.
忽视混杂变量：假设你早上给运动员新能量饮料，下午给安慰剂来测试效果。时间是一个混杂因素；你无法区分任何差异究竟是饮料还是时间造成的。
Insufficient randomisation: Allowing researchers to assign treatments “by convenience” or letting subjects choose introduces selection bias. Always use a recognised random mechanism.
随机化不足：让研究者“由方便”分配处理，或让受试者自选，会引入选择偏倚。始终使用公认的随机机制。
Pseudoreplication: Treating multiple measurements from the same unit as independent replicates leads to artificially small standard errors. If you measure the same plant five times, those five values are not five independent replicates.
假重复：把来自同一个单元的多次测量当作独立的重复，会导致标准误人为偏小。如果你对同一株植物测量五次，这五个值并不是五个独立的重复。
10. Exam-Style Application Tips | 考试应用技巧
When faced with an A-Level question on experimental design, structure your response clearly. Start by identifying the experimental units, the treatments, and the response variable.
面对A-Level中有关实验设计的题目时，要清晰地组织你的答案。首先明确实验单元、处理因素和响应变量。
Then, describe the design in a logical sequence: how randomisation will be carried out, the number of replicates, any blocking or blinding, and how data will be collected. Use phrases like “to minimise bias” and “to ensure that any observed difference is due to the treatments alone.”
然后按逻辑顺序描述设计：如何进行随机化，重复次数，任何区组或盲法措施，以及如何收集数据。使用诸如“以最大程度减少偏倚”和“以确保观察到的任何差异仅归因于处理”这样的短语。
If the question asks for a practical improvement, think about control groups, increasing replicates, or introducing blinding. Always link your suggestion back to the principle it addresses. For example, “Introducing a double-blind protocol would eliminate both subject and experimenter bias, strengthening internal validity.”
如果题目要求提出实际改进，考虑增加对照组、增加重复次数或引入盲法。始终将你的建议与所解决的原则关联起来。例如，“引入双盲方案将消除受试者和实验者偏倚，增强内部效度。”
Finally, comment on limitations: an experiment conducted in a laboratory may lack ecological validity; a small sample size may limit generalisability. Examiners value such reflective conclusions.
最后，评述局限性：在实验室进行的实验可能缺乏生态效度；样本量过小可能限制可推广性。考官们看重这样反思性的总结。
11. Statistical Tools for Analysing Experimental Data | 分析实验数据的统计工具
Once data are collected, we turn to statistical tests. Common tools in A-Level Mathematics include the t-test for comparing two means, chi-squared tests for association, and correlation/regression for relationships between variables.
数据收集完成后，我们转而使用统计检验。A-Level数学中常用的工具包括用于比较两个均值的t检验、用于关联性的卡方检验，以及用于变量间关系的相关与回归分析。
For a simple two-treatment completely randomised experiment, a two-sample t-test (or paired t-test for matched pairs) is appropriate. The test statistic for a two-sample t-test under the assumption of equal variances is:
t = (x̄₁ – x̄₂) / (sₚ √(1/n₁ + 1/n₂))
对于简单的两处理完全随机实验，可采用两样本t检验（配对设计则用配对t检验）。假设方差相等时，两样本t检验的统计量为：
t = (x̄₁ – x̄₂) / (sₚ √(1/n₁ + 1/n₂))
where sₚ is the pooled standard deviation. You should be able to interpret the p-value in context: a small p-value (typically < 0.05) suggests the observed difference is statistically significant.
其中sₚ为合并标准差。你应能够结合上下文解释p值：较小的p值（通常<0.05）表明观察到的差异具有统计显著性。
Always check assumptions before applying a test. For a t-test, the data should be approximately normal and the samples independent. If the experiment involved blocking, a more advanced analysis of variance (ANOVA) might be used, but at A-Level, comparing block-adjusted means often suffices.
在应用检验之前，务必检查假设。对于t检验，数据应近似正态，样本独立。如果实验涉及区组，可能要用更高级的方差分析，但在A-Level水平，比较调整后的区组均值通常就够了。
12. Bringing It All Together: A Worked Example | 综合案例
Imagine a student wants to investigate whether a new revision app improves test scores. She recruits 40 volunteers and randomly assigns 20 to use the app (treatment) and 20 to use traditional notes (control). The test scores out of 50 are recorded. The data yield a sample mean difference of 4.2 with a pooled standard deviation of 6.1. The calculated t-statistic is 2.18 with 38 degrees of freedom, giving a p-value of 0.036. She concludes there is evidence at the 5% significance level that the app improves scores.
设想一位学生想研究一款新的复习应用是否能提高测试成绩。她招募了40名志愿者，随机分配20人使用该应用（处理组），20人使用传统笔记（对照组）。记录满分为50的测试成绩。数据得出样本均值差为4.2，合并标准差为6.1。计算出的t统计量为2.18，自由度为38，p值为0.036。她得出结论，在5%显著性水平下，有证据表明该应用能提高成绩。
However, she reflects that the volunteers might be more motivated than average students, limiting generalisability. She also notes that without blinding, a placebo effect might be at play. To improve, she could add a placebo app group and ensure the test markers do not know group assignments.
然而，她反思志愿者可能比一般学生更有动力，这限制了结论的可推广性。她还注意到，由于缺乏盲法，可能存在安慰剂效应。为了改进，她可以增加一个安慰剂应用组，并确保阅卷人不清楚分组情况。
This worked example touches on randomisation, control, replication, analysis, and critical evaluation – exactly the blend of skills examiners look for.
这个综合案例涉及随机化、对照、重复、分析和批判性评估——正是考官所看重的综合技能。
Published by TutorHao | Mathematics 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