📚 Pre-U AQA Statistics: Case Study Practical Exercises | Pre-U AQA 统计：案例分析实战演练
In Pre-U AQA Statistics, the ability to apply statistical reasoning to real-world scenarios is essential. Case study practical exercises bridge the gap between isolated techniques and holistic data analysis, requiring you to formulate hypotheses, select appropriate tests, justify assumptions, and interpret findings in context. This article walks you through a structured approach to tackling such case studies, using a consistent example that mirrors the depth and complexity expected at this level. By engaging with each stage—from study design to critical evaluation—you will build the confidence to handle any exam-style statistical investigation.
在 Pre-U AQA 统计课程中，将统计推理应用于真实情境的能力至关重要。案例分析实战演练填补了孤立技巧与整体数据分析之间的空白，它要求你提出假设、选择合适的检验方法、论证假设条件，并在具体情境中解读结果。本文将通过一个结构化的方法来引导你解决这类案例，使用一个贯穿始终的示例，反映该级别考试所要求的深度与复杂性。通过参与从研究设计到批判性评价的每一个阶段，你将树立信心，从容应对任何考试风格的统计调查。
1. Understanding the Case Study Framework | 理解案例分析框架
A Pre-U AQA Statistics case study typically presents a short narrative describing a research context, a set of data or experimental information, and a series of questions. Your first task is to identify the population, the variables involved, and the type of study—whether it is observational, experimental, or based on a survey. Pay close attention to the wording: phrases such as ‘random sample’, ‘control group’, or ‘before-after measurements’ signal the statistical design. You should also note the scale of measurement for each variable: nominal, ordinal, interval, or ratio. This initial mapping directly influences which statistical tools are appropriate later.
一份 Pre-U AQA 统计案例通常呈现一段简短的叙述，描述研究背景、一组数据或实验信息，以及一系列问题。你的首要任务是确定总体、涉及的变量以及研究类型——是观察性的、实验性的还是基于调查的。密切关注措辞：像“随机样本”、“对照组”或“前后测量”之类的字眼提示了统计设计。你还应留意每个变量的测量尺度：名义、有序、间隔或比率。这一初步梳理直接决定了后续哪些统计工具是适用的。
2. Defining the Research Question and Hypotheses | 确定研究问题与假设
Once the context is clear, extract the core research question. Suppose a case study investigates whether a new online revision platform improves A-level students’ test scores compared to traditional self-study. The null hypothesis H₀ might state there is no difference in mean scores, while the alternative H₁ could be one-sided (the platform increases mean scores). Always formulate hypotheses in population parameters, not sample statistics. For instance, H₀: μ_new = μ_traditional versus H₁: μ_new > μ_traditional, where μ denotes the population mean test score. Be precise about the parameter of interest: it could be a mean, a proportion, a difference, or an association measure.
一旦厘清背景，就应提炼核心研究问题。假设一个案例调查新的在线复习平台是否比传统自学更能提高 A-level 学生的考试成绩。原假设 H₀ 可能表述为平均分无差异，而备择假设 H₁ 可以是单侧的（该平台提高平均分）。始终用总体参数而非样本统计量来建立假设。例如，H₀: μ_新 = μ_传统 对应 H₁: μ_新 > μ_传统，其中 μ 表示总体平均考试分数。要精确指明关注的参数：它可能是均值、比例、差值或关联度量。
3. Summarising and Displaying the Data | 数据汇总与展示
Before conducting any formal analysis, you must summarise the data with appropriate descriptive statistics and graphics. For the revision platform example, assume two independent groups of 25 students each, with their post-intervention scores recorded. Compute means, medians, standard deviations, and quartiles for each group. Display the distributions using side-by-side boxplots or histograms to visually assess symmetry, outliers, and central tendency. In an exam response, sketch these plots clearly and comment on patterns: e.g., ‘The new platform group appears shifted to the right with less variability, suggesting a potential effect.’ Always label axes and provide a legend if needed.
在进行任何正式分析之前，你必须用恰当的描述性统计量和图形来汇总数据。针对复习平台的例子，假设有两组独立学生，每组 25 人，记录了他们的干预后分数。计算每组的均值、中位数、标准差和四分位数。使用并列箱线图或直方图展示分布，以便直观评估对称性、离群值和集中趋势。在考试作答中，应清晰地画出这些图示并评论模式：例如，“新平台组看起来右移且变异性较小，暗示可能存在效应”。务必标记坐标轴，并在需要时提供图例。
4. Checking Assumptions for Inferential Tests | 检验推断方法的假定条件
A hallmark of Pre-U AQA Statistics is the emphasis on checking assumptions before selecting a test. For a two-sample t-test comparing means, you need approximately normal distributions in each population (or large enough samples) and equal variances unless using Welch’s correction. Assess normality using normal probability plots or the Shapiro-Wilk test if summary statistics are provided. Check equality of variances with an F-test or Levene’s test. If assumptions are violated, consider a non-parametric alternative such as the Mann-Whitney U test. Comments like ‘Given the sample size of 25, the Central Limit Theorem may justify approximate normality, but the boxplots show slight skew, so I will also perform a Mann-Whitney test as a sensitivity analysis’ demonstrate critical thinking.
Pre-U AQA 统计的一个标志是重视在选择检验方法前检查假定条件。对于比较均值的双样本 t 检验，需要每个总体近似正态分布（或样本量足够大），以及方差相等（除非采用 Welch 校正）。利用正态概率图或 Shapiro-Wilk 检验（若提供汇总统计量）评估正态性。用 F 检验或 Levene 检验检查方差齐性。如果假定不满足，可考虑非参数替代方法，如 Mann-Whitney U 检验。像“鉴于样本量为 25，中心极限定理可能支持近似正态，但箱线图显示轻微偏斜，因此我也会执行 Mann-Whitney 检验作为敏感性分析”这样的评述能展现批判性思维。
5. Performing the Hypothesis Test | 执行假设检验
Carry out the chosen test step by step. For an independent samples t-test, calculate the test statistic t = (x̄₁ – x̄₂) / (s_p √(1/n₁ + 1/n₂)), where s_p is the pooled standard deviation. Determine the degrees of freedom, find the critical value or p-value from statistical tables, and compare with a significance level α (commonly 0.05). State your decision clearly: ‘Since p = 0.018 < 0.05, we reject H₀ and conclude there is sufficient evidence to suggest the new platform increases mean test scores.' If using a non-parametric test, outline the ranking procedure and report the U-statistic. Include a comment on the practical significance: a statistically significant result may still reflect a trivially small effect size, so compute Cohen's d or the Hodges-Lehmann estimator.
逐步执行所选的检验。对于独立样本 t 检验，计算检验统计量 t = (x̄₁ – x̄₂) / (s_p √(1/n₁ + 1/n₂))，其中 s_p 为合并标准差。确定自由度，从统计表中查找临界值或 p 值，并与显著性水平 α（通常为 0.05）比较。明确陈述你的决定：“由于 p = 0.018 < 0.05，我们拒绝 H₀，并认为有充分证据表明新平台提高了平均考试分数。”若使用非参数检验，概述排序过程并报告 U 统计量。还应包含对实际显著性的评论：统计显著的结果可能对应的效应量很小，因此要计算 Cohen's d 或 Hodges-Lehmann 估计量。
6. Dealing with Categorical Data: Chi-Squared Tests | 处理分类数据：卡方检验
Many case studies involve categorical variables. Suppose the same revision platform study also recorded whether each student rated the resources as ‘helpful’, ‘neutral’, or ‘not helpful’. You could test if there is an association between group (platform vs. traditional) and rating using a chi-squared test of independence. The test statistic is χ² = Σ (Oᵢ – Eᵢ)² / Eᵢ, where Oᵢ are observed frequencies and Eᵢ are expected under independence. Degrees of freedom = (rows – 1) × (columns – 1). Remember to check that no expected frequency is less than 1 and that at most 20% are below 5; otherwise, combine categories or use Fisher’s exact test. Interpret the result: rejection implies the rating distribution differs by group.
许多案例研究涉及分类变量。假设同一个复习平台研究还记录了每位学生是否将资源评为“有帮助”、“中立”或“无帮助”。你可以使用独立性卡方检验来检验组别（平台 vs. 传统）与评分之间是否存在关联。检验统计量为 χ² = Σ (Oᵢ – Eᵢ)² / Eᵢ，其中 Oᵢ 为观测频数，Eᵢ 为独立性假设下的期望频数。自由度 = (行数 – 1) × (列数 – 1)。切记检查期望频数不得小于 1，且低于 5 的比例不超过 20%；否则需合并类别或使用 Fisher 精确检验。解释结果：拒绝原假设意味着评分分布因组别而异。
7. Correlation and Regression | 相关与回归
When the research question concerns the relationship between two numerical variables—for example, hours spent on the platform and final score—scatter plots and correlation coefficients are the starting point. Compute Pearson’s r to measure linear association. However, r can be misleading if outliers are present or the relationship is non-linear, so always supplement with a scatter plot. If a linear model is appropriate, fit a least-squares regression line y = a + b x. Interpret the slope b: ‘For each additional hour of platform use, the test score is predicted to increase by b points, on average.’ Assess the fit using the coefficient of determination R², and check residuals for constant variance and independence.
当研究问题涉及两个数值变量之间的关系时——例如在平台上花费的小时数与最终分数——散点图和相关系数是起点。计算 Pearson 积差相关系数 r 以度量线性关联。然而，如果存在离群值或关系是非线性的，r 可能产生误导，因此始终辅以散点图。若线性模型合适，则拟合最小二乘回归线 y = a + b x。解释斜率 b：“平台使用时间每增加一小时，考试分数预计平均提高 b 分。”用决定系数 R² 评估拟合优度，并检查残差的方差稳定性和独立性。
8. Confidence Intervals and Uncertainty | 置信区间与不确定性
Hypothesis tests alone are not enough; Pre-U AQA Statistics expects you to construct and interpret confidence intervals (CIs). A 95% CI for the difference in means (μ₁ – μ₂) gives a plausible range of the true effect size. For the t-test example, the interval might be (1.2, 5.6). Since the entire interval lies above zero, it reinforces the conclusion that the new platform leads to higher scores, and it quantifies the benefit. For regression, construct CIs for the slope or a prediction interval for a new observation. Always link the CI back to the context: ‘We are 95% confident that the true mean score improvement is between 1.2 and 5.6 marks.’
仅有假设检验是不够的；Pre-U AQA 统计要求你构建并解释置信区间。均值差 (μ₁ – μ₂) 的 95% 置信区间给出了真实效应大小的合理范围。对于 t 检验示例，区间可能为 (1.2, 5.6)。由于整个区间位于零值上方，这加强了新平台带来更高分数的结论，并量化了效益。对于回归，构建斜率的置信区间或新观测值的预测区间。始终将置信区间与情境联系起来：“我们有 95% 的把握认为真实平均分数提高在 1.2 至 5.6 分之间。”
9. Assumptions and Limitations: Critical Evaluation | 假设与局限：批判性评价
A top-tier answer addresses the study’s limitations. Discuss potential sources of bias: was the sample truly random and representative of the target population? In the platform example, volunteer bias might occur if more motivated students opted for the new platform. Consider confounding variables: students using the platform might also complete more practice tests. Mention the limitation of p-values, the effect of sample size, and the risk of Type I and Type II errors. If assumptions were borderline, reflect on how robust your conclusions are. This evaluation shows a mature statistical understanding beyond mechanical computation.
一份顶尖的答案会探讨研究的局限性。讨论潜在的偏差来源：样本是否真正随机并代表目标总体？在平台示例中，如果动机更强的学生选择了新平台，就可能出现志愿者偏差。考虑混杂变量：使用平台的学生可能也完成了更多的练习测试。提及 p 值的局限性、样本量的影响以及第一类和第二类错误的风险。如果假定条件处于临界状态，应反思结论的稳健性。这种评价展现了超越机械计算的成熟统计理解。
10. Integrating Multiple Statistical Methods | 整合多种统计方法
Real case studies often require you to synthesise several techniques. You might begin with a chi-squared test on categorical user feedback, then move to a t-test for comparing average scores, and finally a regression analysis linking usage time to scores. Show how the results complement each other. For instance, the chi-squared test reveals a significant difference in satisfaction, the t-test confirms a mean score advantage, and regression quantifies the dose-response relationship. Present your findings in a coherent narrative, perhaps using a table to summarise key test statistics and p-values. This integration mirrors the analytical reporting expected in professional statistics.
真实的案例研究往往要求你综合运用多种技术。你可以从分类用户反馈的卡方检验开始，然后进行 t 检验以比较平均分数，最后通过回归分析将使用时间与分数联系起来。展示这些结果如何相互印证。例如，卡方检验显示满意度存在显著差异，t 检验证实了平均分数优势，回归则量化了剂量反应关系。以连贯的叙述呈现你的发现，或许使用一个表格汇总关键检验统计量和 p 值。这种整合体现了专业统计所期望的分析报告能力。
11. Practice Case Study with Worked Solution | 带详细解答的实战案例
Let us work through a concise case. Scenario: A school investigates whether a ‘flipped classroom’ approach affects A-level Statistics exam performance. Two classes of 30 students are randomly assigned: Class A uses the flipped model, Class B traditional lectures. The end-of-topic test scores (out of 50) are summarised as Class A: mean = 36.4, s = 6.2; Class B: mean = 32.8, s = 7.1. Additionally, students rated their confidence on a 3-point scale (low, medium, high). Tasks: (a) Conduct an appropriate hypothesis test for the difference in means, using α = 0.05. (b) The confidence rating data are given as a 2×3 contingency table, with observed frequencies: for flipped class, low 5, medium 10, high 15; for traditional class, low 12, medium 11, high 7. Perform a chi-squared test at α = 0.05 to determine if confidence rating is associated with teaching method. (c) Provide a brief critical evaluation.
让我们逐步完成一个简练的案例。情境：一所学校调查“翻转课堂”模式是否影响 A-level 统计考试成绩。两个班级各 30 名学生随机分配：A 班采用翻转模式，B 班采用传统授课。单元测试成绩（满分 50 分）汇总为：A 班：均值 = 36.4, s = 6.2；B 班：均值 = 32.8, s = 7.1。此外，学生按 3 点量表（低、中、高）评定了自信心。任务：(a) 进行适当的假设检验以判断均值差异，α = 0.05。(b) 信心评级数据以一个 2×3 列联表给出，观测频数为：翻转班，低 5、中 10、高 15；传统班，低 12、中 11、高 7。执行 α = 0.05 的卡方检验，判断信心评级是否与教学方式相关。(c) 提供简短的批判性评价。
Solution (a): Hypotheses: H₀: μ_A = μ_B, H₁: μ_A ≠ μ_B (two-tailed). Assumptions: independence of groups, approximate normality (samples of 30 justify CLT). Check equal variances: F = s_B² / s_A² = 7.1²/6.2² ≈ 50.41/38.44 ≈ 1.31, with df (29,29); critical F at 5% (two-tailed) ≈ 2.10, so we assume equal variances. Pooled s_p = √((29×6.2² + 29×7.1²)/58) = √((1115.56+1461.89)/58) = √(2577.45/58) ≈ √44.44 ≈ 6.67. Test statistic t = (36.4 – 32.8) / (6.67 √(1/30+1/30)) = 3.6 / (6.67 × √0.0667) = 3.6 / (6.67×0.2582) ≈ 3.6 / 1.722 ≈ 2.09. df = 58, p-value for two-tailed test ≈ 0.041 (from tables). Since p < 0.05, reject H₀. There is evidence of a difference in mean scores, with the flipped class scoring higher.
解答 (a)：假设：H₀: μ_A = μ_B，H₁: μ_A ≠ μ_B（双侧）。假定：组间独立，近似正态（样本量 30 支持中心极限定理）。检验等方差：F = s_B² / s_A² = 7.1²/6.2² ≈ 1.31，自由度 (29,29)；5% 双侧临界 F ≈ 2.10，因此我们假设方差相等。合并标准差 s_p = √((29×6.2² + 29×7.1²)/58) ≈ 6.67。检验统计量 t = (36.4 – 32.8) / (6.67 √(1/30+1/30)) ≈ 2.09。自由度 58，双侧 p 值 ≈ 0.041。由于 p < 0.05，拒绝 H₀。有证据表明平均分存在差异，翻转班分数更高。
Solution (b): Table of observed and expected (in brackets):
| Low | Medium | High | Total | |
| Flipped | 5 (8.5) | 10 (10.5) | 15 (11.0) | 30 |
| Traditional | 12 (8.5) | 11 (10.5) | 7 (11.0) | 30 |
| Total | 17 | 21 | 22 | 60 |
Expected values all above 5, so chi-squared test is appropriate. χ² = (5-8.5)²/8.5 + (10-10.5)²/10.5 + (15-11)²/11 + (12-8.5)²/8.5 + (11-10.5)²/10.5 + (7-11)²/11 = 1.441 + 0.024 + 1.455 + 1.441 + 0.024 + 1.455 = 5.84 (approx). df = (2-1)×(3-1) = 2. Critical value at 5% is 5.991, so χ² = 5.84 < 5.991, we fail to reject H₀. At the 5% significance level, there is insufficient evidence of an association between teaching method and confidence rating.
解答 (b)：观测与期望（括号内）表见上。期望值均大于 5，适合卡方检验。χ² ≈ 5.84，自由度 2，5% 临界值 5.991，未能拒绝 H₀。在 5% 显著性水平下，尚无充分证据表明教学方式与信心评级之间存在关联。
Solution (c): The t-test suggests improved performance in the flipped class, but the study is small and randomisation may not control for teacher effect. The chi-squared test is non-significant but close to the critical value; a larger sample might reveal a difference. We should also consider effect sizes: Cohen’s d = (36.4-32.8)/√((6.2²+7.1²)/2) ≈ 3.6/6.67 ≈ 0.54, a moderate effect. Future studies should account for baseline ability and use matched pairs.
解答 (c)：t 检验表明翻转班成绩有所提高，但研究规模较小，随机化可能无法控制教师效应。卡方检验不显著但接近临界值；更大的样本可能揭示差异。我们还应考虑效应量：Cohen’s d ≈ 0.54，为中等效应。未来研究应控制基线能力并使用配对设计。
12. Common Pitfalls and Exam Tips | 常见陷阱与应试技巧
When working through case studies, students often rush into testing without properly defining hypotheses or checking assumptions. Always annotate your steps. Avoid using a t-test when the data are paired or when variances are grossly unequal without adjustment. In chi-squared tests, remember that the test is two-tailed by nature. Interpret p-values correctly: they are not the probability that H₀ is true. Label all graphs, quote the significance level, and provide a conclusion in the context of the original problem. Finally, leave a few minutes for a final review to ensure your answers form a logically coherent report. Practice with timed case studies under exam conditions to build fluency.
在处理案例分析时，学生常常匆忙进入检验阶段，而未能正确定义假设或检查假定条件。始终为你的步骤加上注解。数据成对或方差异常悬殊且未做调整时，应避免使用 t 检验。在卡方检验中，务必记住该检验本质上是双侧的。正确解读 p 值：它们并非原假设为真的概率。为所有图形添加标签，标明显著性水平，并在原始问题的背景下给出结论。最后，留出几分钟进行最终检查，确保答案构成一份逻辑连贯的报告。在考试条件下限时练习案例分析，以培养流畅性。
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