Pre-U Cambridge Statistics: Practical/Experimental Exam Key Points | Pre-U Cambridge 统计：实验/实践考核要点

📚 Pre-U Cambridge Statistics: Practical/Experimental Exam Key Points | Pre-U Cambridge 统计：实验/实践考核要点

The Pre-U Cambridge Statistics practical examination is a distinctive component that evaluates not just theoretical knowledge, but the ability to apply statistical methods to real-world data. Success requires mastering experimental design, data handling, choosing appropriate tests, and presenting findings effectively. This guide distils the essential key points to help you excel in the practical assessment.

Pre-U 剑桥统计实践考试是一个独特的组成部分，它不仅评估理论知识，还考察将统计方法应用于真实数据的能力。成功需要掌握实验设计、数据处理、选择合适的检验方法以及有效地呈现结果。本指南提炼了重要的关键要点，帮助你在此实践评估中脱颖而出。

1. Understanding the Practical Exam Structure | 理解实践考试的结构

Familiarity with the exam format is your first step. The practical paper typically presents one or more data-based scenarios, often derived from scientific investigations, surveys, or simulations. You will be asked to explore data, conduct tests, check assumptions, and write concise conclusions. Time management is critical, as marks are distributed across data manipulation, analysis, and interpretation.

熟悉考试形式是你的第一步。实践试卷通常会给出一个或多个基于数据的场景，这些场景通常来自科学研究、调查或模拟。你将被要求探索数据、进行检验、检查假设条件，并撰写简洁的结论。时间管理至关重要，因为分数分配在数据处理、分析和解释上。

Scenarios may involve comparing groups, testing associations, or modelling relationships.
场景可能涉及组别比较、关联性检验或关系建模。
You must use statistical software (such as Excel, R, or a calculator with statistical functions) but all working and decision-making must be clearly documented.
你必须使用统计软件（如 Excel、R 或带有统计功能的计算器），但所有的计算过程和决策必须清晰地记录下来。

2. Mastering Experimental Design | 掌握实验设计

Before touching any data, think like an experimenter. The exam often expects you to suggest improvements or critique a given design. Keywords such as randomisation, blocking, control groups, and replication will earn you marks. Explain why randomisation helps eliminate bias and how blocking accounts for nuisance variation.

在接触任何数据之前，像实验者一样思考。考试通常希望你提出改进建议或对给定的设计进行评判。如随机化、区组、对照组和重复等关键词会为你赢得分数。解释为什么随机化有助于消除偏差，以及区组如何解释讨厌变量。

Ideally, treatments should be assigned randomly to experimental units.
理想情况下，处理应随机分配给实验单元。
Replication increases the reliability of the estimate of experimental error and the power of the test.
重复提高了对实验误差估计的可靠性以及检验的功效。

3. Data Collection and Sampling Techniques | 数据收集与抽样技术

You may be given raw data and asked to comment on the sampling method. Distinguish between simple random sampling, stratified sampling, systematic sampling, and cluster sampling. Identify potential sources of bias, such as non-response or convenience sampling, and state how they affect the validity of conclusions.

你可能会得到原始数据，并被要求对抽样方法进行评论。区分简单随机抽样、分层抽样、系统抽样和整群抽样。识别潜在的偏差来源，如无回应或便利抽样，并说明它们如何影响结论的有效性。

Stratified sampling ensures representation of subgroups; it often increases precision.
分层抽样确保子群体的代表性；它通常能提高精确度。
A convenience sample cannot be used to make valid inferences about a population.
便利样本无法用于对总体做出有效推断。

4. Data Cleaning and Preparation | 数据清洗与准备

Real data is messy. In the practical exam, you will frequently need to identify outliers, missing values, and incorrect entries. State when an outlier should be investigated rather than automatically removed. For missing data, consider whether it is missing completely at random (MCAR) and note that imputation or omission may affect results.

真实数据是杂乱的。在实践考试中，你经常需要识别异常值、缺失值和错误输入。说明何时应调查异常值而不是自动删除。对于缺失数据，考虑它是否完全随机缺失 (MCAR)，并注意插补或删除可能会影响结果。

Always keep a log of any data changes you make, explaining your reasoning.

始终记录你所做的任何数据更改，并解释你的理由。

5. Exploratory Data Analysis: Visual Storytelling | 探索性数据分析：用图表讲故事

Before formal hypothesis testing, always visualise. Boxplots reveal medians, spread, and potential outliers; histograms and density plots show the shape of the distribution. For relationships, use scatterplots. For categorical data, consider bar charts or mosaic plots. Labelling axes, providing titles, and describing what the plot suggests are indispensable.

在进行正式的假设检验之前，始终要进行可视化。箱线图揭示中位数、分散程度和潜在的异常值；直方图和密度图显示分布的形状。对于关系，使用散点图。对于分类数据，考虑使用条形图或马赛克图。标注坐标轴、提供标题并描述图表所揭示的信息是不可或缺的。

Comment on skewness, multimodality, and gaps – these diagnostics guide test choice.
评论偏度、多峰和间隙——这些诊断指导检验的选择。
If you see a non-linear pattern in a scatterplot, transforming the data or using a non-parametric test may be appropriate.
如果在散点图中看到非线性模式，对数据进行转换或使用非参数检验可能是合适的。

6. Choosing the Right Statistical Test | 选择正确的统计检验

This is a pivotal skill. Your decision tree should follow the data type and research question. Is the response variable continuous or categorical? How many groups are being compared? Are the samples independent or paired? Is the normality assumption plausible? These questions lead you to t-tests, ANOVA, chi-square tests, or non-parametric alternatives like Mann-Whitney and Wilcoxon.

这是一项关键技能。你的决策树应依据数据类型和研究问题。响应变量是连续的还是分类的？比较多少个组？样本是独立的还是配对的？正态性假设是否合理？这些问题会引导你选择 t 检验、方差分析、卡方检验或像 Mann-Whitney 和 Wilcoxon 这样的非参数替代方法。

Study Goal / 研究目标	Parametric Test / 参数检验	Non-parametric / 非参数
Compare two independent groups / 比较两独立组	Independent t-test	Mann-Whitney U
Compare two paired groups / 比较两配对组	Paired t-test	Wilcoxon signed-rank
Compare three or more independent groups / 比较三组及以上独立组	One-way ANOVA	Kruskal-Wallis
Association between two categorical variables / 两分类变量关联	Chi-square test of independence

7. Performing Hypothesis Tests Correctly | 正确地执行假设检验

Always state your null (H₀) and alternative (H₁) hypotheses clearly. Define the significance level α, usually 0.05. Compute the test statistic using software but verify the formula conceptually. Report the p-value accurately: e.g. p = 0.032 rather than ‘p < 0.05'. Never say 'accept H₀'; instead state 'there is insufficient evidence to reject H₀'.

始终清晰地陈述你的零假设 (H₀) 和备择假设 (H₁)。定义显著性水平 α，通常是 0.05。使用软件计算检验统计量，但在概念上验证公式。准确地报告 p 值：例如 p = 0.032，而不是 ‘p < 0.05'。永远不要说 '接受 H₀'；而应说 '没有足够证据拒绝 H₀'。

If conducting a two-tailed test, ensure you double the one-tailed p-value where appropriate, and justify your choice of tails based on the research question.

如果进行双尾检验，确保在适当时将单尾 p 值加倍，并根据研究问题证明你选择尾数的理由。

8. Interpreting p-values and Confidence Intervals | 解读 p 值与置信区间

A p-value is the probability of obtaining test results at least as extreme as the observed, assuming H₀ is true. It is not the probability that H₀ is true. Emphasise the link to confidence intervals: a 95% confidence interval for a difference that does not include zero corresponds to a significant result at α = 0.05 (two-sided). Provide practical significance alongside statistical significance; a very small effect might be statistically significant with a huge sample but have no real-world relevance.

p 值是在 H₀ 为真的前提下，获得至少与实际观测一样极端的检验结果的概率。它不是 H₀ 为真的概率。强调与置信区间的联系：差异的 95% 置信区间如果不包含零，则对应于在 α = 0.05（双侧）下显著。将实际显著性与统计显著性一起提供；一个非常小的效应在样本量巨大时可能统计显著，但没有现实意义。

If μ₁ – μ₂ 95% CI = (0.5, 2.8), we are 95% confident that μ₁ exceeds μ₂ by between 0.5 and 2.8 units.

如果 μ₁ – μ₂ 的 95% 置信区间 = (0.5, 2.8)，我们有 95% 的把握认为 μ₁ 比 μ₂ 高 0.5 到 2.8 个单位。

9. Checking Assumptions Thoroughly | 彻底检查假设条件

Every parametric test rests on assumptions. For t-tests and ANOVA: independence, approximate normality, and homogeneity of variances. Use residual plots to assess these. A normal probability plot (Q-Q plot) helps detect non-normality; a plot of residuals versus fitted values shows whether variance is constant. If assumptions are severely violated, either transform the data (e.g., log, square root) or select an alternative non-parametric test. Document this decision.

每个参数检验都基于假设条件。对于 t 检验和方差分析：独立性、近似正态性和方差齐性。使用残差图来评估这些假设。正态概率图 (Q-Q 图) 有助于发现非正态性；残差与拟合值的图显示了方差是否恒定。如果假设条件严重违反，要么转换数据（例如，对数、平方根），要么选择替代的非参数检验。记录这个决定。

For chi-square tests, verify expected frequencies are >= 5; combine categories if necessary.
对于卡方检验，验证期望频数 >= 5；如有必要，合并类别。
For correlation, check for a linear relationship and absence of influential outliers.
对于相关分析，检查线性关系和没有强影响点。

10. Drawing Conclusions and Structuring Your Report | 得出结论并构建你的报告

The exam requires a structured, concise report. Begin with a brief objective, then outline the methodology: design, data description, and chosen analysis. Present descriptive statistics and relevant graphs. Clearly show the test statistic, degrees of freedom, p-value, and effect size if asked. End with a conclusion that directly answers the research question, acknowledges limitations, and suggests how the experiment could be improved.

考试要求一份结构清晰、简洁的报告。以一个简短的目标开始，然后概述方法：设计、数据描述和选择的分析。呈现描述性统计和相关图表。清楚地显示检验统计量、自由度、p 值，以及如果要求，效应大小。以一个直接回答研究问题、承认局限性并建议如何改进实验的结论结束。

Use bullet points or numbered sections if the question instructs, but maintain academic tone. Avoid flowery language; precision is valued.

如果问题指示，使用项目符号或编号部分，但保持学术语气。避免华丽的语言；精确性受到重视。

11. Common Pitfalls and How to Avoid Them | 常见陷阱及如何避免

Many students lose marks by jumping to a paired test when the design is unpaired, or overlooking that a ‘significant’ result does not prove causation. Also, failing to examine raw data can hide data entry errors or odd patterns. Running multiple tests without adjusting significance level (e.g., Bonferroni correction) inflates the type I error rate. Mention this if you perform post-hoc comparisons after ANOVA.

许多学生因为设计是非配对的却跳到配对检验，或者忽略了 ‘显著’ 结果并不能证明因果关系而失分。此外，未能检查原始数据可能会隐藏数据输入错误或奇怪模式。在不调整显著性水平（例如，Bonferroni 校正）的情况下进行多次检验会增大 I 类错误率。如果在方差分析后进行事后比较，请提及这一点。

Over-reliance on p-values: integrate confidence intervals and effect sizes.
过度依赖 p 值：结合置信区间和效应大小。
Forgetting to state assumptions: even if the test is performed by software, you must verify them manually or via plots.
忘记陈述假设条件：即使检验由软件执行，你也必须手动或通过图表验证它们。

12. Tips for Success in the Practical Examination | 实践考试的成功秘诀

Practice with diverse datasets before the exam. Familiarity with the software’s output interpretation saves precious minutes. Prepare a checklist: 1. Understand the context; 2. Clean and explore data; 3. Check assumptions and select test; 4. Conduct analysis; 5. Interpret and conclude. During the exam, read the question carefully – sometimes it specifies the test to use or requires a critical commentary. Allocate time for each stage and leave a few minutes to review your report for clarity and completeness.

在考试前用多样化的数据集进行练习。熟悉软件输出解读可以节省宝贵的时间。准备一个检查清单：1. 理解背景；2. 清理和探索数据；3. 检查假设并选择检验；4. 进行分析；5. 解读并得出结论。考试期间，仔细阅读问题——有时会指定要使用的检验或要求进行批判性评论。为每个阶段分配时间，并留出几分钟回顾报告，确保清晰和完整。

Remember, the examiner is looking for evidence of sound statistical reasoning, not just number crunching. Your explanation of why you chose a certain test and what the results really mean carries the highest weight.

请记住，考官寻找的是合理的统计推理证据，而不仅仅是数字运算。你对于为什么选择某个检验以及结果真正意味着什么的解释，权重最高。

Published by TutorHao | Statistics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Pre-U Cambridge Statistics: Practical/Experimental Exam Key Points | Pre-U Cambridge 统计：实验/实践考核要点