📚 Edexcel Year 13 Statistics: Report Writing Framework and Model Essays | 爱德思13年级统计：报告写作框架与范文
The Edexcel A Level Statistics specification (9ST0) includes a substantial statistical enquiry project that counts towards the final grade. Mastering the report structure is essential for achieving high marks. This guide breaks down a proven framework and provides model paragraphs to help you write a compelling and academically rigorous report.
爱德思 A Level 统计学（9ST0）包含一项计入最终成绩的重要统计调查项目。掌握报告结构对于获得高分至关重要。本指南分解了一套经过验证的框架，并提供范文段落，帮助你撰写令人信服且学术严谨的报告。
Everything below is organised around the six stages of the investigative cycle, the way examiners actually mark it, and the exact wording that separates a top-band report from a mid-band one.
以下内容围绕调查周期的六个阶段、考官实际的评分方式，以及区分高分段与中等分段报告的具体措辞来组织。
1. Understanding the Edexcel Statistical Enquiry | 理解爱德思统计调查项目
The statistical enquiry requires you to demonstrate the entire investigative cycle: posing a problem, planning, collecting data, processing and presenting data, interpreting results, and evaluating the process. The project is assessed against criteria that reward clear communication, correct statistical methods, and critical reflection.
统计调查要求你展示完整的调查周期：提出问题、制定计划、收集数据、处理和呈现数据、解释结果以及评估过程。项目评分标准奖励清晰的表达、正确的统计方法和批判性反思。
Your report should read like a coherent narrative, not a disjointed set of calculations. Maintain a formal academic tone throughout and use consistent statistical notation — if you define x̄ as the sample mean on the first page, keep that symbol everywhere, and never switch to "average" mid-report.
你的报告应读起来像连贯的叙述，而不是零散的计算。始终保持正式的学术语气，并使用一致的统计符号——如果你在第一页把 x̄ 定义为样本均值，那么全文都要沿用这个符号，不能中途改称"平均数"。
The six stages of the cycle map directly onto the sections of your report, which is why the framework below is not arbitrary: it mirrors how the mark scheme is structured.
调查周期的六个阶段与报告的各个部分一一对应，因此下面的框架并非随意设计：它直接反映了评分方案的结构。
| Stage 阶段 | What you do 你要做的事 | Report section 报告位置 |
| Problem 问题 | Define variables and hypotheses 定义变量与假设 | Introduction 引言 |
| Plan 计划 | Sampling, timeline, ethics 抽样、时间表、伦理 | Methodology 方法 |
| Data 数据 | Collect, clean, code 收集、清洗、编码 | Data collection 数据收集 |
| Analysis 分析 | Summary statistics and tests 统计量与检验 | Analysis 分析 |
| Interpretation 解释 | Link findings to hypotheses 将结论与假设关联 | Conclusion 结论 |
| Evaluation 评估 | Critique limitations 批判局限性 | Evaluation 评估 |
2. Selecting a Research Question | 选择研究问题
Choose a question that is grounded in real-world context and can be answered using statistical methods within the time and resources you realistically have. A good enquiry question names the population, the variables, and the relationship or difference you intend to investigate.
选择一个有真实背景、并且在你实际拥有的时间和资源内可以用统计方法回答的问题。一个好的调查问题会明确指出总体、变量，以及你打算研究的关联或差异。
Weak questions are vague ("Is social media bad?"), unmeasurable ("Is maths harder than physics?"), or impossible to resource ("How does income affect happiness across Europe?"). Strong questions are specific, measurable, and moderate in scope.
弱问题通常含糊（"社交媒体有害吗？"）、无法测量（"数学比物理难吗？"），或资源上不可行（"收入如何影响全欧洲的幸福感？"）。好问题则具体、可测量、范围适中。
Your hypotheses must be stated formally and before you look at the data, so that the test you choose later genuinely tests the claim rather than confirming it.
你的假设必须正式陈述，并且在看数据之前就确定，这样后续选择的检验才是真正在检验该命题，而不是在事后迎合它。
H₀: ρ = 0 H₁: ρ ≠ 0 (α = 0.05, two-tailed)
- State the null and alternative hypotheses in symbols as well as words. 用符号和文字同时陈述原假设与备择假设。
- Justify a one-tailed or two-tailed test from your context. 根据背景论证单尾或双尾检验的合理性。
- Fix the significance level in advance (α = 0.05 or α = 0.01) and commit to it. 提前确定显著性水平并坚持使用。
3. Planning and Project Management | 规划与项目管理
A convincing methodology section shows that your plan was deliberate. Explain where the data came from, how many observations you aimed for, how long collection would take, and how you would handle non-responses.
有说服力的方法部分要显示你的计划是经过深思熟虑的。说明数据来源、目标观测数量、收集所需时间，以及你如何处理无应答情况。
Ethical and practical considerations carry marks: informed consent, anonymity, age-appropriate questions, data protection, and permission to access school records. Mentioning these explicitly is not padding — it is evidence of professional planning.
伦理与实践考量是可以得分的：知情同意、匿名性、问题适合年龄、数据保护，以及访问学校记录的许可。明确提及这些并非凑字数——这正是专业规划的体现。
A simple Gantt-style timeline demonstrates feasibility and gives you something concrete to reflect on in the evaluation.
一个简单的甘特图式时间表既能展示可行性，也为你在评估部分提供具体的反思素材。
| Week 周次 | Activity 活动 | Output 产出 |
| 1-2 | Literature scan, question refinement 文献浏览、问题打磨 | Final question 最终问题 |
| 3 | Pilot questionnaire 预调查问卷 | Revised instrument 修订后的工具 |
| 4-5 | Data collection 数据收集 | Raw dataset 原始数据集 |
| 6-7 | Cleaning and analysis 清洗与分析 | Charts, test statistics 图表与检验统计量 |
| 8 | Writing and proofreading 写作与校对 | Final report 最终报告 |
4. Data Collection and Sampling | 数据收集与抽样
Distinguish clearly between primary data, which you collect yourself for this specific question, and secondary data, which already exists. Each has trade-offs you must acknowledge.
要清楚区分一手数据（你为这个问题亲自收集的）和二手数据（已经存在的）。两者各有取舍，你必须予以说明。
Name your sampling method precisely: simple random, systematic, stratified, cluster, quota, or opportunity. Then state what bias each method introduces. Opportunity sampling is convenient but rarely representative; stratified sampling improves precision but requires reliable population information.
要准确说出你的抽样方法：简单随机、系统、分层、整群、配额或便利抽样。然后说明每种方法会引入什么偏差。便利抽样方便但很少具代表性；分层抽样提高精度，但需要可靠的总体信息。
Sample size should be justified, not guessed. For a target margin of error E at confidence level 1 − α, the required size for a mean is:
样本量应当被论证，而不是凭感觉。在置信水平 1 − α、目标误差界限为 E 时，估计均值所需的样本量为：
n = (z × σ ÷ E)²
For a proportion use the same structure with p(1 − p) in place of σ². Where σ is unknown, use a pilot study estimate and say so explicitly.
估计比例时使用同样的结构，只需把 σ² 换成 p(1 − p)。若 σ 未知，可用预调查的估计值，并在报告中明确说明。
- Report the achieved sample size and the response rate separately. 分别报告实际样本量和应答率。
- Explain how non-response might bias the estimate. 解释无应答如何使估计产生偏差。
- Note any variable you could not measure and why. 说明无法测量的变量及其原因。
5. Data Processing and Cleaning | 数据处理与清洗
Before any analysis, describe how raw responses became a usable dataset. Examiners look for evidence of cleaning decisions being made consciously and recorded.
在做任何分析之前，先说明原始回答如何变成可用数据集。考官希望看到清洗决策是有意识做出并被记录的。
Outliers are a decision point, not an automatic deletion. Either keep them and comment on their effect on the mean, or remove them with a stated statistical justification and show the result both ways.
异常值是一个决策点，而非自动删除项。要么保留它们并说明其对均值的影响，要么在有明确统计理由的情况下剔除，并把两种结果都呈现出来。
Missing data should be described, not silently dropped. Report how many cases were affected and whether the missingness is likely to be random.
缺失数据应当被描述，而不是悄悄丢弃。报告受影响的数据条数，以及缺失是否可能是随机的。
Outlier rule: value < Q₁ − 1.5 × IQR or value > Q₃ + 1.5 × IQR
Standardising units before comparison avoids a common error: comparing a mass in grams with a mass in kilograms, or a duration in minutes with one in seconds.
在比较前统一单位可以避免一个常见错误：把以克为单位的质量与以千克为单位的质量相比，或把以分钟计的时长与以秒计的时长相比。
6. Exploratory Analysis and Presentation | 探索性分析与呈现
Every chart in your report must earn its place by answering part of the question. A histogram shows distribution shape; a box plot compares groups and exposes skewness and outliers; a scatter graph shows the form, direction and strength of an association.
报告中的每张图表都必须通过回答问题的某一部分来证明其价值。直方图展示分布形状；箱线图比较组别并揭示偏态与异常值；散点图展示关联的形态、方向和强度。
Always pair the visual with the numerical summary. A scatter graph without the product moment correlation coefficient is descriptive; the coefficient plus the graph is analytical.
始终把图形与数值摘要配对。只有散点图是描述性的；加上积矩相关系数才是分析性的。
| Variable type 变量类型 | Best display 最佳呈现 | Summary statistics 摘要统计量 |
| Continuous, one group 连续型单组 | Histogram 直方图 | x̄, s, median, IQR |
| Continuous, two groups 连续型两组 | Back-to-back box plots 背靠背箱线图 | Medians, quartiles, ranges |
| Bivariate 双变量 | Scatter graph 散点图 | PMCC, regression line |
| Categorical vs categorical 分类对分类 | Contingency table 列联表 | Observed and expected counts |
Use the standard deviation s where the distribution is reasonably symmetric and the interquartile range where it is skewed. 当分布大致对称时用标准差 s，偏斜时用四分位距。
7. Choosing and Applying Statistical Tests | 选择与应用统计检验
Test selection is driven by the data type and the question being asked, not by which test you find easiest. Justify the choice in one or two sentences and the examiner immediately sees statistical reasoning.
检验的选择由数据类型和研究问题决定，而不是由你觉得哪种检验最简单决定。用一两句话论证选择，考官立刻就能看到统计推理能力。
For association between two continuous variables, use the product moment correlation coefficient when the relationship is roughly linear, and Spearman's rank correlation coefficient when it is monotonic but curved or when the data are ordinal.
对于两个连续变量之间的关联，当关系大致线性时使用积矩相关系数；当关系单调但呈曲线，或数据为有序变量时，使用斯皮尔曼等级相关系数。
The test statistics you will most commonly need are shown below, with their degrees of freedom.
你最常需要的检验统计量及其自由度如下所示。
r = Σ(xᵢ − x̄)(yᵢ − ȳ) ÷ √[ Σ(xᵢ − x̄)² × Σ(yᵢ − ȳ)² ]
t = (x̄ − μ₀) ÷ (s ÷ √n) with ν = n − 1
χ² = Σ (Oᵢ − Eᵢ)² ÷ Eᵢ with ν = (r − 1)(c − 1)
- State H₀ and H₁, the test, the significance level, the critical value, the calculated statistic and the conclusion. 依次陈述原假设与备择假设、所用检验、显著性水平、临界值、计算得到的统计量与结论。
- Write the comparison explicitly: for example, 0.412 > 0.306, so reject H₀. 明确写出比较过程：例如 0.412 > 0.306，因此拒绝 H₀。
- Translate the conclusion back into context in a full sentence. 用一个完整句子把结论翻译回实际情境。
8. Regression and Correlation Modelling | 回归与相关建模
Correlation measures the strength of an association; regression produces a model you can use for prediction. Correlating does not prove causation, and your report should say so explicitly rather than implying it.
相关衡量关联的强度；回归产生可用于预测的模型。相关并不证明因果，你的报告应当明确说明这一点，而不是暗示存在因果。
The least squares regression line of y on x is found from the summary statistics, and the line always passes through the point (x̄, ȳ).
y 对 x 的最小二乘回归直线可由摘要统计量求出，并且该直线必定经过点 (x̄, ȳ)。
y = a + bx where b = Sₓᵧ ÷ Sₓₓ and a = ȳ − b x̄
for each one-unit increase in x, and the intercept as the predicted value of y when x = 0 — but only if x = 0 lies inside the data range and makes physical sense.
把斜率解释为 x 每增加一个单位时 y 的预测变化量，把截距解释为 x = 0 时 y 的预测值——但前提是 x = 0 落在数据范围内且具有现实意义。
- Use interpolation inside the observed range; flag extrapolation as unreliable. 在观测范围内使用内插；把外推标记为不可靠。
- Report the coefficient of determination, r², as the proportion of variation explained. 报告决定系数 r²，即被解释的变异比例。
- Check residual plots for random scatter before trusting the linear model. 在信任线性模型之前，检查残差图是否随机散布。
9. Writing the Analysis and Interpretation | 撰写分析与解释
The analysis section presents results; the interpretation section explains what they mean for your original question. Keeping them distinct prevents the common fault of a report that describes numbers without ever answering anything.
分析部分呈现结果；解释部分说明这些结果对你的原始问题意味着什么。把两者分开，可以避免"只描述数字却从不回答问题"这个常见缺陷。
Every statistical conclusion must be paired with a contextual conclusion. "r = 0.78" is a statistic; " on hours and test score, so students who revise more tended to score higher in this sample" is an answer.
每一个统计结论都必须和一个情境结论配对。"r = 0.78" 只是一个统计量；"有强证据表明每周复习时长与测验分数之间存在正相关，因此在本样本中复习更多的学生得分往往更高"才是答案。
Report uncertainty honestly. Use a confidence interval alongside the point estimate, and interpret the p-value correctly: it is the probability of observing data at least this extreme if H₀ were true, not the probability that H₀ is true.
诚实地报告不确定性。在点估计之外给出置信区间，并正确解释 p 值：它是在 H₀ 成立的前提下，观测到至少如此极端数据的概率，而不是 H₀ 为真的概率。
x̄ ± z × (σ ÷ √n) or x̄ ± t × (s ÷ √n)
10. Evaluation and Critical Reflection | 评估与批判性反思
The evaluation is where mid-band reports stall and top-band reports are made. Generic statements such as "the sample could have been bigger" earn little; a specific, quantified critique earns a lot.
评估部分正是中等分段报告止步、高分段报告脱颖而出之处。"样本本可以更大"这类泛泛之谈得分很低；具体、有量化支撑的批评才能得高分。
Structure your evaluation around three questions: how valid are the results, how reliable are they, and how could the study be improved if repeated?
围绕三个问题组织评估：结果的效度如何、信度如何，以及如果重做这项研究可以如何改进？
| Issue 问题 | Weak comment 弱评论 | Strong comment 强评论 |
| Sample size 样本量 | Too small 太小 | With n = 18 the 95% confidence interval spans 12.4 marks, so the estimate lacks precision. 当 n = 18 时，95% 置信区间宽达 12.4 分，估计精度不足。 |
| Bias 偏差 | It was biased 有偏差 | Opportunity sampling in one sixth form over-represents higher-attaining students, inflating the mean. 在单一六年级学院使用便利抽样会过度代表高成就学生，从而抬高均值。 |
| Measurement 测量 | Data may be wrong 数据可能有误 | Self-reported revision hours are subject to recall bias, likely towards round numbers. 自报复习时长存在回忆偏差，且倾向取整数。 |
End with a concrete improvement: a stratified sample of 60 students, a two-week diary rather than a single recall question, and a repeat test with an independent sample to check reliability.
最后给出具体的改进建议：对 60 名学生进行分层抽样，采用为期两周的记录表而非单一回忆题，并用独立样本重复检验以核对信度。
11. Model Paragraphs and Mark-Winning Phrases | 范文段落与高分句式
Below are three model paragraphs written in the register examiners expect. Adapt the structure rather than copying the content, since your own context will differ.
以下是三段符合考官期待语体的范文段落。请借鉴其结构而非照抄内容，因为你的具体情境不同。
Model introduction 范文引言
ds on independent study per week and their score on a 40-mark mathematics mock examination. The population is the 214 Year 13 students at the college; the sample consists of 45 students selected by stratified random sampling across four mathematics classes. The null hypothesis is H₀: ρ = 0 and the alternative is H₁: ρ > 0, tested at the 5% significance level using Spearman's rank correlation coefficient.
本调查旨在确定 13 年级学生每周独立学习时数与其 40 分数学模拟考试成绩之间是否存在显著正相关。总体为该学院 214 名 13 年级学生；样本为通过分层随机抽样从四个数学班中选出的 45 名学生。原假设为 H₀: ρ = 0，备择假设为 H₁: ρ > 0，使用斯皮尔曼等级相关系数在 5% 显著性水平下检验。
Model analysis paragraph 范文分析段
duct moment correlation coefficient is r = 0.612. With n = 45, the critical value at the 5% o reject H₀. The coefficient of determination r² = 0.375 indicates that approximately 37.5% ng the majority attributable to other factors such as prior attainment and teaching group.
散点图显示中等程度的正线性关联，没有明显的异常值。积矩相关系数为 r = 0.612。当 n = 45 时，5% 水平下单尾检验的临界值为 0.248；由于 0.612 > 0.248，有充分证据拒绝 H₀。决定系数 r² = 0.375 表明，模拟成绩约 37.5% 的变异可由学习时数的变异解释，其余大部分归因于先前学业水平和教学班等其他因素。
Model evaluation paragraph 范文评估段
The study is limited by its reliance on self-reported study hours, which are vulnerable to recall bias and social desirability bias. As the sample was drawn from a single college, external validity is restricted and the findings should not be generalised to the national Year 13 population. The correlation also does not establish causation: students with higher prior attainment may both study more and score higher. A stronger design would pair each student's self-report with a two-week study diary and repeat the measurement in a second institution.
本研究的局限在于依赖自报学习时数
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图