A-Level WJEC Mathematics: Practical Guide to Experimental Design | A-Level WJEC数学：实验设计实用指南

📚 A-Level WJEC Mathematics: Practical Guide to Experimental Design | A-Level WJEC数学：实验设计实用指南

Experiments form the backbone of evidence-based decision-making in science, industry and everyday life. Whether you are testing a new drug, comparing fertilizer yields or evaluating a teaching method, a well-planned experiment allows you to draw valid conclusions about cause and effect. In the WJEC A-Level Mathematics specification, particularly within the Statistics component, you are expected to understand the core principles of experimental design and to be able to critique and construct simple experiments. This guide walks you through the key concepts, practical steps and common pitfalls so you can master experimental design with confidence.

实验是科学、工业乃至日常生活中循证决策的基石。无论是测试新药、比较肥料产量还是评估教学方法，一个精心设计的实验能让你得出关于因果关系的有效结论。在 WJEC A-Level 数学中，尤其是统计学部分，你需要理解实验设计的核心原则，并能批判和构建简单的实验。本指南将带你梳理关键概念、实操步骤和常见误区，帮助你自信掌握实验设计。

1. What Is an Experiment? | 什么是实验？

An experiment is a controlled study in which the researcher deliberately imposes a treatment on individuals (called experimental units) to observe the effect on a response variable. Unlike an observational study, an experiment actively intervenes, making it possible to establish causation rather than mere association. For example, giving half of a class a new revision technique and the other half the standard method is an experiment; simply surveying students about their revision habits is an observational study.

实验是一种受控研究，研究者有目的地对个体（称为实验单位）施加某种处理，观察其对响应变量的影响。与观察性研究不同，实验主动干预，从而能够确立因果关系而不仅仅是关联。例如，让一半学生使用新的复习方法，另一半使用标准方法，这就是实验；而仅仅调查学生的复习习惯就是观察性研究。

The hallmark of a true experiment is random assignment of treatments to units. This minimises bias and helps ensure that any observed differences in the response can be attributed to the treatment rather than to lurking variables. In WJEC questions, you will often need to distinguish between an experiment and other types of study and justify why an experiment is needed to infer causation.

真正实验的标志是将处理随机分配给单位。这能将偏差降到最低，并有助于确保观察到的响应差异可归因于处理，而不是混杂变量。在 WJEC 试题中，你经常需要区分实验与其他类型的研究，并论证为何需要实验才能推断因果关系。

2. Experimental Units and Treatments | 实验单位与处理

Experimental units are the individual entities to which a treatment is applied. They could be people, animals, plots of land, electronic components, or any other discrete item. Treatments are the specific conditions or interventions being tested. In a simple comparative experiment, there might be just two treatments: a new drug (treatment) and a placebo (control). Sometimes treatments can be combinations of factors, such as temperature and pressure in an industrial process.

实验单位是接受处理的个体对象。它们可以是人、动物、地块、电子元件或任何离散的物体。处理是被测试的具体条件或干预措施。在一个简单的比较实验中，可能只有两种处理：新药（处理组）和安慰剂（对照组）。有时处理可以是因子的组合，比如工业过程中的温度和压力。

When defining units and treatments, clarity is crucial. For instance, if you want to test the effect of a fertilizer on wheat yield, each plot of land is an experimental unit, and the different fertilizer types or amounts are the treatments. In WJEC exam scenarios, you may be asked to identify the units, treatments and response variable from a description of a study.

在定义单位和处理时，清晰明确至关重要。例如，如果你想测试肥料对小麦产量的影响，每块土地就是一个实验单位，而不同的肥料类型或用量就是处理。在 WJEC 考试情境中，你可能会被要求从研究描述中识别出单位、处理和响应变量。

3. Response Variable and Measurement | 响应变量与测量

The response variable is the outcome that is measured on each experimental unit after the treatment has been applied. It should be measured in a consistent, accurate, and unbiased manner. In the wheat example, the response variable could be the weight of grain harvested per plot. In a drug trial, it might be the reduction in blood pressure. A good response variable is directly relevant to the research question and is as objective as possible.

响应变量是在施加处理后，对每个实验单位测量的结果。它应以一致、准确和无偏的方式进行测量。在小麦例子中，响应变量可以是每个地块收获的谷物重量。在药物试验中，可能是血压的降低值。一个好的响应变量应直接与研究问题相关，并尽可能客观。

Measurement errors can ruin an otherwise well-designed experiment. It is important to use the same measuring instruments, protocols, and timing for all units to avoid systematic bias. In WJEC, you might discuss the reliability and validity of measurements and suggest improvements, such as using double-blind procedures or calibrated equipment.

测量误差会毁掉一个原本设计良好的实验。对所有单位使用相同的测量仪器、方案和时间点以避免系统偏差非常重要。在 WJEC 中，你可能会讨论测量的信度和效度，并提出改进建议，比如使用双盲程序或校准设备。

4. Principles of Experimental Design: Randomisation | 实验设计原则：随机化

Randomisation means using a chance process to assign treatments to experimental units. This ensures that each unit has an equal probability of receiving any treatment, which tends to balance out the effects of confounding variables that you may not even be aware of. Without randomisation, systematic differences between groups could mask or exaggerate treatment effects.

随机化意味着使用随机过程将处理分配给实验单位。这确保了每个单位有相同的概率接受任何一种处理，从而往往能平衡掉你可能都没意识到的混杂变量的影响。没有随机化，组间的系统性差异可能掩盖或夸大处理效应。

Simple randomisation can be achieved by flipping a coin, using random number tables, or computer-generated random sequences. In WJEC, you should be able to explain why randomisation is essential and suggest how to implement it in a given context. For example, 'assign 30 patients a number and use random digits to allocate 15 to the drug group and 15 to the placebo group'.

简单的随机化可以通过抛硬币、使用随机数表或计算机生成的随机序列来实现。在 WJEC 中，你应能解释为何随机化必不可少，并能在特定情境下建议如何实施。例如，“给 30 名病人编号，用随机数字将 15 人分配到药物组，15 人到安慰剂组”。

5. Principles of Experimental Design: Replication | 实验设计原则：重复

Replication is the application of each treatment to multiple experimental units. It serves two main purposes: it allows you to estimate the natural variability in the response, and it increases the precision of treatment effect estimates. If you apply a treatment to only one unit, you cannot tell whether the result is typical or just an anomaly. Replication provides the data needed for formal statistical tests like the t-test or ANOVA.

重复是将每种处理施加到多个实验单位上。它有两个主要目的：让你能估计响应的自然变异性，并提高处理效应估计的精度。如果每种处理只施加在一个单位上，你就无法判断结果是典型还是异常。重复提供了进行正式统计检验（如 t 检验或方差分析）所需的数据。

In practice, replication means having enough units in each treatment group. The required sample size depends on the expected effect size and variability. In WJEC problems, you might be asked to comment on whether an experiment has adequate replication. A study with only two patients per group is almost useless, whereas 30 per group is often considered a reasonable minimum for many contexts.

在实践中，重复意味着每个处理组有足够的单位。所需样本量取决于预期的效应大小和变异性。在 WJEC 问题中，你可能会被要求评论一个实验是否有足够的重复。一个每组只有两名病人的研究几乎毫无用处，而每组 30 人在许多情境下通常被视为合理的最低限度。

6. Principles of Experimental Design: Control and Blocking | 实验设计原则：对照与区组

A control group serves as a baseline for comparison. It receives no treatment, a placebo, or the standard treatment. Without a control, you cannot determine what would have happened in the absence of the experimental treatment. Blocking is a technique used to account for known sources of variability among experimental units. Units are first divided into blocks such that units within a block are similar, and then treatments are randomly assigned within each block.

对照组用于作为比较的基线。它不接受处理、接受安慰剂或标准处理。没有对照，你就无法确定在没有实验处理的情况下会发生什么。区组化是一种用于解释实验单位之间已知变异来源的技术。先将单位划分为区组，使得区组内的单位相似，然后在每个区组内随机分配处理。

For example, in a crop experiment, you might block by field location to account for soil fertility gradients. Each block would contain each treatment once. Blocking reduces the impact of the blocked variable on the estimated treatment effect, making the experiment more sensitive. WJEC questions often ask you to design a block layout or explain why blocking is beneficial.

广告 / Sponsored

例如，在一个作物实验中，你可以按田间位置划分区组以考虑土壤肥力梯度。每个区组包含每种处理一次。区组化减少了被区组变量对处理效应估计的影响，使实验更灵敏。WJEC 问题经常要求你设计一个区组布局或解释区组化的好处。

7. Blinding and the Placebo Effect | 盲法与安慰剂效应

Blinding means preventing the subjects, those administering the treatment, or those assessing the response from knowing which treatment a unit received. A single-blind experiment keeps the subjects unaware; a double-blind experiment keeps both subjects and researchers unaware. This reduces bias arising from psychological expectations or subtle cues.

盲法意味着防止受试者、实施处理者或评估响应者知道某个单位接受了哪种处理。单盲实验让受试者不知道；双盲实验让受试者和研究者都不知道。这减少了来自心理预期或微妙暗示的偏差。

The placebo effect is a genuine improvement in a patient's condition simply because they believe they are receiving an effective treatment. A placebo is an inert substance or procedure that mimics the treatment but has no therapeutic effect. By comparing treatment and placebo groups under blind conditions, you can separate the true pharmacological effect from the psychological one. In WJEC, you need to recognise when blinding or a placebo is appropriate and explain its role.

安慰剂效应是指病人仅仅因为相信自己正在接受有效治疗而出现的真实病情改善。安慰剂是一种模拟治疗但没有疗效的惰性物质或程序。通过在盲法条件下比较处理组和安慰剂组，你可以分离出真实的药理效应和心理效应。在 WJEC 中，你需要识别何时适合采用盲法或安慰剂，并解释其作用。

8. Completely Randomised Design (CRD) | 完全随机化设计

A completely randomised design is the simplest experimental design, where all experimental units are assigned to treatments entirely at random, with no blocking. It assumes that the units are homogeneous, or at least that any differences will be averaged out by randomisation. This design is easy to implement and analyse, typically using a two-sample t-test or one-way ANOVA for more than two treatments.

完全随机化设计是最简单的实验设计，其中所有实验单位完全随机地分配到各个处理，没有区组。它假设单位是同质的，或者至少能通过随机化将任何差异平均掉。这种设计易于实施和分析，通常使用双样本 t 检验或针对两个以上处理的单因素方差分析。

For the CRD to be effective, randomisation must be genuine. In WJEC, you might be given a list of units and asked to allocate them randomly to treatments using a method of your choice. You must also be able to state the advantages (simplicity) and disadvantages (not efficient if there is known variation among units) of a completely randomised design.

为了让完全随机化设计有效，随机化必须是真实的。在 WJEC 中，你可能会被给定一组单位，并要求用你选择的方法将它们随机分配到处理组。你还必须能够说出完全随机化设计的优点（简单）和缺点（如果单位间存在已知变异则效率不高）。

9. Randomised Block Design (RBD) | 随机化区组设计

In a randomised block design, experimental units are first grouped into blocks based on a characteristic that is expected to affect the response (e.g., age, soil type, initial blood pressure). Within each block, treatments are randomly assigned to units. This ensures that each treatment is fairly compared under similar conditions, removing block-to-block variability from the experimental error.

在随机化区组设计中，首先根据一个预期会影响响应的特征（如年龄、土壤类型、初始血压）将实验单位分组为区组。在每个区组内，处理被随机分配给单位。这确保了每种处理都在相似条件下进行公平比较，将区组间的变异性从实验误差中移除。

Analysis of an RBD often uses a two-way ANOVA without interaction, where the block is another factor. The key advantage is increased precision for the treatment comparisons if the blocking factor is indeed influential. In WJEC, you should be able to recognise when blocking is appropriate and draw a diagram showing the layout. A typical exam task is: 'Design a randomised block experiment to compare three types of battery, using four different devices as blocks.'

随机化区组设计的分析通常使用无交互作用的双因素方差分析，其中区组是另一个因子。如果区组因子确实有影响，主要优势是提高处理比较的精度。在 WJEC 中，你应能识别何时适合区组化，并画出展示布局的图表。典型的考试任务是：“设计一个随机化区组实验来比较三种电池，用四种不同的设备作为区组。”

10. Matched Pairs Design | 配对设计

A matched pairs design is a special case of blocking where each block consists of only two similar units. The two units are then randomly assigned to the two treatments (or one to treatment and one to control). This design is particularly useful when experimental units are scarce or when you can pair them logically, such as twins, left-right arms, or pre-test and post-test scores on the same person.

配对设计是区组化的一种特殊情况，其中每个区组只包含两个相似的单位。这两个单位随后被随机分配到两种处理（或一个到处理组，一个到对照组）。当实验单位稀缺或可以逻辑上配对时，比如双胞胎、左右臂，或同一个人的前测和后测分数，这种设计特别有用。

Data analysis for matched pairs involves calculating the difference within each pair and then performing a one-sample t-test on the differences (or a paired t-test). This eliminates between-subject variability, leading to a more powerful test. WJEC questions sometimes ask you to explain why a matched pairs design is more appropriate than a completely randomised design for a given scenario, such as comparing the effectiveness of a moisturising cream on left and right hands.

配对设计的数据分析涉及计算每对内部的差值，然后对差值进行单样本 t 检验（或配对 t 检验）。这消除了受试者之间的变异，使检验更有效力。WJEC 问题有时会问，为什么在给定情境下配对设计比完全随机化设计更合适，比如比较保湿霜在左手和右手上的效果。

11. Analysing Data from Experiments | 分析实验数据

ments from independent groups, a two-sample t-test is commonly used. The null hypothesis (H₀) typically states that there is no difference between the treatment means (μ₁ = μ₂), while the alternative (H₁) suggests a difference. The test produces a p-value; a small p-value (usually p < 0.05) provides evidence against H₀.

收集数据后，统计分析有助于判断观察到的差异是真实的还是偶然造成的。对于比较两个独立组的处理，通常使用双样本 t 检验。原假设（H₀）通常表述为处理均值之间没有差异（μ₁ = μ₂），备择假设（H₁）则表明存在差异。检验产生 p 值；较小的 p 值（通常 p < 0.05）提供反对 H₀ 的证据。

For more than two treatments, one-way ANOVA compares means and checks whether at least one group mean differs significantly. In WJEC, you are not required to compute ANOVA by hand but you need to interpret output, state hypotheses, and relate results to the experimental context. It is also important to consider confidence intervals for the difference between means to assess the practical significance, not just statistical significance.

对于两个以上的处理，单因素方差分析比较均值并检查是否至少有一组均值存在显著差异。在 WJEC 中，你不需要手动计算方差分析，但需要解释输出结果、陈述假设，并将结果与实验背景联系起来。考虑均值差异的置信区间也很重要，以评估实际意义，而不仅仅是统计显著性。

12. Critiquing an Experimental Design | 批判实验设计

A core skill tested by WJEC is the ability to evaluate a described experiment and identify strengths and weaknesses. You should systematically check: Is there random allocation? Is there a control group? Is replication sufficient? Have potential sources of bias (measurement, selection, confounding) been addressed? Is blinding used where necessary? Could results be generalised to the population of interest?

WJEC 考核的一项核心技能是评估所描述的实验，并指出优点和缺点。你应系统性地检查：是否有随机分配？是否有对照组？重复是否充分？潜在的偏倚来源（测量、选择、混杂）是否得到解决？必要时是否使用了盲法？结果能否推广到感兴趣的人群？

Common flaws in student-designed experiments include small sample sizes, lack of a control, non-random assignment, and failure to account for obvious confounding variables like time of day or operator skill. When critiquing, always suggest concrete improvements. For instance, 'The experiment does not use a placebo, so the placebo effect cannot be ruled out. A double-blind, placebo-controlled design would strengthen the validity significantly.'

学生设计的实验中常见的缺陷包括样本量小、缺少对照、非随机分配，以及未能考虑明显的混杂变量，如一天中的时间或操作者技能。在进行批判时，总是要提出具体的改进建议。例如，“该实验未使用安慰剂，因此无法排除安慰剂效应。采用双盲、安慰剂对照的设计将显著增强有效性。”

Published by TutorHao | A-Level WJEC Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）