📚 Large Data Sets: Key Points & Analysis Methods | 大数据集考点与分析方法
The large data set (LDS) is a compulsory component of A-Level Mathematics Statistics. Exam boards such as Edexcel, AQA and OCR provide a pre-released dataset, and you must be familiar with its context, variables and typical exam-style questions. This article consolidates the key points and analytical methods you need to master.
大数据集是 A-Level 数学统计部分的必考内容。Edexcel、AQA、OCR 等考试委员会都会提供预先公布的数据集，考生必须熟悉其背景、变量以及典型考题。本文将系统梳理你需要掌握的核心考点与分析方法。
1. What Is a Large Data Set? | 什么是大数据集
A large data set is a substantial collection of real-world data — typically over 1,000 observations — used to teach statistical reasoning beyond small textbook examples. In the exam, you are expected to recall facts about the dataset without being given the data sheet itself.
大数据集是真实世界数据的较大规模集合——通常超过 1000 条观测记录——用于训练超越教材小样本的统计推理能力。考试中，你需要在没有数据表的情况下回忆数据集的具体信息。
Edexcel: Met Office weather data covering 5 UK sites (Camborne, Leeming, Hurn, Shawbury, Heathrow) plus 3 overseas islands (Lerwick, Stornoway, Ringway? Actually these are also UK — Edexcel uses 5 UK weather stations: Camborne, Leeming, Hurn, Shawbury, Heathrow) over 1987–2015, with variables such as daily maximum temperature, rainfall, wind speed, and pressure.
Edexcel：英国气象局气象数据，涵盖 5 个英国站点（Camborne、Leeming、Hurn、Shawbury、Heathrow），时间跨度为 1987–2015 年，变量包括每日最高气温、降雨量、风速和气压等。
AQA: a large data set based on cars (make, age, price, engine size, CO₂ emissions) and a health dataset on children's growth and lifestyle.
AQA：基于汽车的数据集（品牌、车龄、价格、发动机排量、CO₂ 排放量）以及儿童成长与生活方式健康数据集。
OCR: a dataset covering both weather (similar to Edexcel) and transport/speed data on trains.
OCR：数据集同时涵盖气象数据（与 Edexcel 相似）和铁路运输/速度数据。
2. Types of Data | 数据类型
You must distinguish between qualitative and quantitative data, and between discrete and continuous data. Weather variables such as temperature and wind speed are continuous; daily rainfall totals are continuous but often recorded to one decimal place; the number of rainy days is discrete.
你必须区分定性数据与定量数据，以及离散数据与连续数据。气温、风速等气象变量属连续数据；每日降雨总量是连续的，但通常保留一位小数；雨天日数则是离散数据。
Data Type Decision Chain: Questions → Words/Counts? → Discrete or Continuous?
数据类型判断链：问题 → 文字/计数？→ 离散还是连续？
Qualitative (qualitative 定性): categories such as wind direction ("SW", "NE"). No arithmetic is meaningful.
定性：类别性记录，如风向（"西南"、"东北"），对其进行算术运算无实际意义。
Quantitative (定量): numerical data. Discrete (离散) data can only take specific values, e.g. number of sunny days; continuous (连续) data can take any value in an interval, e.g. temperature 15.2 °C.
定量：数值型数据。离散数据只能取特定值，如晴天日数；连续数据可在区间内取任意值，如气温 15.2 °C。
3. Measures of Location | 位置度量
Measures of location summarise the central tendency or positional values of a dataset. The three most important are the mean, median and mode; for large data sets you also need quartiles and percentiles.
位置度量用于概括数据集的中心趋势或位置特征。最重要的三个指标是平均数、中位数和众数；对于大数据集，还需要掌握四分位数和百分位数。
Mean: x̄ = Σx ⁄ n | 平均数：x̄ = Σx ⁄ n
For grouped data use midpoints: x̄ = Σfx ⁄ Σf. Choose the correct class interval carefully.
分组数据使用组中值：x̄ = Σfx ⁄ Σf。选择正确的组区间至关重要。
The median Q₂ is the middle value when data are ordered. For n observations, Q₂ is the (n+1)/2-th value. Interpolation may be required for grouped data.
中位数 Q₂ 是有序数据中的中间值。对于 n 个观测值，Q₂ 是第 (n+1)/2 个值。分组数据可能需要插值。
The lower quartile Q₁ cuts off the bottom 25%; the upper quartile Q₃ cuts off the top 25%. Percentiles divide data into 100 equal parts — the 90th percentile, for example, exceeds 90% of observations.
下四分位数 Q₁ 截断底部 25%；上四分位数 Q₃ 截断顶部 25%。百分位数将数据分成 100 等份——例如第 90 百分位数超过 90% 的观测值。
4. Measures of Spread | 离散程度度量
Measures of spread describe how variable or dispersed the data are. Range and interquartile range (IQR) are easy to interpret, but variance and standard deviation are algebraically more powerful and are required for regression and normal distribution work.
离散程度度量描述数据的变异性或分散程度。极差和四分位距（IQR）易于解释，但方差和标准差在代数上更具优势，是回归和正态分布计算的基础。
Variance = Σ(x − x̄)² ⁄ n = Σx² ⁄ n − x̄² | 方差 = Σ(x − x̄)² ⁄ n = Σx² ⁄ n − x̄²
Standard deviation σ is the positive square root of the variance; it shares the units of the original data, making it directly interpretable.
标准差 σ 是方差的算术平方根；它与原始数据同单位，因此可直接解释。
For grouped data, use Σfx² ⁄ Σf − x̄². Remember to weight by frequency f.
分组数据使用 Σfx² ⁄ Σf − x̄²。注意用频数 f 加权。
QR = Q₃ − Q₁ is robust to outliers, unlike the range which is sensitive to extreme values.
IQR = Q₃ − Q₁ 对异常值稳健，而极差对极端值非常敏感。
5. Graphical Representations | 图表表示
Different graph types serve different purposes. Histograms for continuous data, box plots for comparing distributions, and cumulative frequency graphs for quartiles are the most frequently examined.
不同类型的图表各有用途。直方图用于连续数据，箱线图用于比较分布，累积频率图用于求四分位数——这些是考查频率最高的图表。
| Graph 图表 | Uses 用途 | Key Formula 关键公式 |
| Histogram 直方图 | Continuous data, unequal class widths 连续数据、不等组距 | Frequency density = frequency ÷ class width 频率密度 = 频率 ÷ 组距 |
| Box plot 箱线图 | Compare distributions 比较分布 | Median, Q₁, Q₃, whiskers 中位数、Q₁、Q₃、触须 |
| Cumulative frequency 累积频率图 | Find quartiles and percentiles 求四分位数和百分位数 | Plot cumulative frequency vs upper boundary 以累积频率对上边界作图 |
In a histogram, the area of each bar is proportional to frequency. If class widths differ, you must calculate frequency density before drawing.
在直方图中，每个矩形的面积与频率成正比。若组距不同，必须先计算频率密度再作图。
Box plots show five-number summary: minimum, Q₁, median, Q₃, maximum. Outliers are plotted as individual points beyond the whiskers.
箱线图展示五数概括：最小值、Q₁、中位数、Q₃、最大值。异常值在触须之外以单独的点绘出。
Cumulative frequency curves allow reading of median and quartiles directly from the graph; always plot at the upper class boundary.
累积频率曲线可直接从图中读出中位数和四分位数；作图时务必使用组上边界。
6. Outliers and Cleaning Data | 异常值与数据清洗
Real-world large data sets contain errors, missing values and extreme observations. Exam questions often ask you to clean data, justify the removal of outliers, and explain how missing values should be handled.
真实世界的大数据集包含错误、缺失值和极端观测。考试常让你清洗数据、为删除异常值论证，并解释缺失值的处理方法。
₁ − 1.5 × IQR or above Q₃ + 1.5 × IQR. Some boards prefer 2 × IQR or 3 standard deviations — always match the definition given in the question.
常用异常值定义：小于 Q₁ − 1.5 × IQR 或大于 Q₃ + 1.5 × IQR 的值。有些考试委员会采用 2 × IQR 或 3 个标准差——务必与题目给出的定义保持一致。
When an outlier is identified, you may either remove it (if it is a measurement error) or retain it (if it is a genuine rare event). Justification must reference the context.
识别出异常值后，可以选择删除（若属于测量误差）或保留（若是真实罕见事件）。论证必须结合数据背景。
Missing values should not be silently ignored; state whether you used listwise deletion, interpolation, or the median substitution.
缺失值不应被静默忽略；应说明你采用成列删除、插值还是中位数替代法。
7. Correlation and Regression | 相关与回归
Correlation measures the strength and direction of a linear relationship between two variables. The product-moment correlation coefficient (PMCC) r lies between −1 and +1. Regression finds the line of best fit for prediction.
相关衡量两个变量之间线性关系的强度和方向。皮尔逊积矩相关系数（PMCC）r 介于 −1 与 +1 之间。回归用于找到最佳拟合线以进行预测。
r = Σ(x − x̄)(y − ȳ) ⁄ √(Σ(x − x̄)² Σ(y − ȳ)²)
rfect positive linear correlation; r = −1: perfect negative; r = 0: no linear correlation.
r = +1：完全正线性相关；r = −1：完全负线性相关；r = 0：无线性相关。
Regression line: y = a + bx, where b = Σxy − n·x̄·ȳ ⁄ (Σx² − n·x̄²), and a = ȳ − b·x̄.
回归直线：y = a + bx，其中 b = Σxy − n·x̄·ȳ ⁄ (Σx² − n·x̄²)，且 a = ȳ − b·x̄。
Use the regression line for interpolation (predicting within the range of data) confidently; extrapolation (outside the range) is risky and must be flagged.
回归直线可用于插值（在数据范围内预测），结果较可靠；外推（超出数据范围）存在风险，必须注明。
8. Time Series and Trends | 时间序列与趋势
Large weather datasets are time series — observations recorded at successive times. Analysis focuses on identifying trend, seasonal variation and residual noise.
大型气象数据集属于时间序列——按时间顺序记录的观测值。分析重点在于识别趋势、季节性波动和残差噪声。
Moving averages smooth out seasonal effects. For monthly data, a 12-point moving average removes the annual cycle; for quarterly data, a 4-point moving average is used.
移动平均可平滑季节性影响。对于月度数据，12 点移动平均消除年度周期；对于季度数据，使用 4 点移动平均。
The trend line is drawn through the moving-average points. Seasonal variation is estimated as the difference between the actual value and the trend value.
趋势线穿过移动平均点绘制。季节性波动通过实际值与趋势值的差值估计。
Examiners expect you to describe the trend (e.g. "temperatures generally increased over the 30-year period") and identify the months with maximum/minimum seasonal effect.
考官期望你描述趋势（如"30 年间气温总体上升"）并识别季节性效应最大/最小的月份。
9. Coding and Transformations | 编码与数据变换
constant. If y = (x − a) ⁄ b, then the mean and standard deviation transform predictably.
编码通过减去常数或除以常数来简化计算。若 y = (x − a) ⁄ b，则平均数与标准差会按可预测的方式变换。
ȳ = (x̄ − a) ⁄ b and σ(y) = σ(x) ⁄ |b|
ȳ = (x̄ − a) ⁄ b； σ(y) = σ(x) ⁄ |b|
Adding or subtracting a constant shifts all measures of location but does not affect spread.
加减常数会平移所有位置度量，但不影响离散程度。
Multiplying or dividing by a positive constant scales both location and spread. A negative divisor flips the order and changes the sign of the mean but not the standard deviation's magnitude.
乘以或除以正常数会同比例缩放位置与离散度量。负除数会改变顺序和平均数符号，但标准差的数值不变。
Coding is especially useful for large data sets with awkward numbers, e.g. temperature recorded in tenths of a degree.
编码对包含棘手数值的大数据集尤其有用，例如以十分之一度记录的气温。
10. Using Technology Effectively | 高效使用技术工具
In the exam you must interpret calculator outputs and spreadsheet summaries. You will not be asked to perform heavy computation by hand, but you must know which button gives which statistic.
考试中你需要解读计算器输出和电子表格汇总结果。不会要求手工进行大量计算，但你必须知道哪个按键对应哪个统计量。
On a Casio ClassWiz or similar, the STAT menu gives x̄, σₙ, σₙ₋₁, Σx and Σx² directly. Know the difference between σₙ (population) and σₙ₋₁ (sample) — A-Level Edexcel uses σₙ for the large data set.
在 Casio ClassWiz 或类似计算器上，STAT 菜单直接给出 x̄、σₙ、σₙ₋₁、Σx 和 Σx²。要区分 σₙ（总体）与 σₙ₋₁（样本）——Edexcel A-Level 大数据集使用 σₙ。
For two-variable statistics, the calculator outputs a, b, r and Σxy. Use these directly for regression and PMCC questions.
对于双变量统计，计算器输出 a、b、r 和 Σxy。在回归和 PMCC 题目中直接使用这些值。
Spreadsheets such as Excel use functions: AVERAGE, MEDIAN, STDEV.P, STDEV.S, CORREL, and LINEST. Be ready to interpret printed outputs.
Excel 等电子表格使用函数：AVERAGE、MEDIAN、STDEV.P、STDEV.S、CORREL 和 LINEST。要能够解读打印输出的结果。
11. Common Exam Traps | 常见考试陷阱
Many students lose marks not from weak statistics but from small process errors. The following pitfalls appear repeatedly in examiner reports.
许多学生失分并非因为统计功底弱，而是由于过程上的小错误。以下是考官报告中反复出现的陷阱。
Forgetting units: temperature in °C, wind speed in knots, pressure in hPa. Always write units in your final answer.
忘记单位：气温单位 °C，风速单位节（knots），气压单位 hPa。最终答案务必写出单位。
Confusing class boundaries with class midpoints when computing the mean of grouped data.
计算分组数据平均数时混淆组边界与组中值。
Using the range instead of the standard deviation when asked for "measure of spread" in a dataset with outliers — the standard deviation or IQR is more appropriate.
当数据含异常值而题目要求"离散程度度量"时，使用极差而非标准差或 IQR——后者更为合适。
Misreading the variable definition — e.g. daily maximum gust vs daily mean wind speed are different variables in the Met Office dataset.
误读变量定义——例如气象局数据集中"每日最大阵风"与"每日平均风速"是两个不同变量。
Extrapolating a regression line outside the range without comment — always state the limitation.
未加说明地在数据范围外进行回归外推——务必指出其局限性。
12. Final Revision Checklist | 考前复习清单
Use this checklist to confirm you are fully prepared for any large data set question.
使用这份清单确认你对大数据集题目已做好充分准备。
I can state the source, time period, locations and all variables of my board's large data set.
我能说出所在考试委员会大数据集的来源、时间跨度、站点和所有变量。
I can distinguish data types and choose the correct measure of location and spread.
我能区分数据类型，并选择正确的位置度量与离散度量。
I can draw and interpret histograms, box plots and cumulative frequency graphs.
我能绘制并解读直方图、箱线图和累积频率图。
I can identify and justify outliers using the 1.5 × IQR rule.
我能使用 1.5 × IQR 规则识别异常值并给出论证。
I can calculate and interpret PMCC r and the least-squares regression line.
我能计算并解读 PMCC r 和最小二乘回归线。
I can analyse time series using moving averages and describe trend/seasonality.
我能用移动平均分析时间序列，并描述趋势与季节性。
I can apply coding transforms correctly and interpret calculator outputs.
我能正确应用编码变换并解读计算器输出。
Mastering the large data set is not about memorising every number — it is about understanding the context, choosing the right tool, and interpreting results in the real-world setting. Practice with past paper questions on your board's specific dataset, and you will turn this compulsory component into a reliable source of marks.
掌握大数据集的关键不在于记住每一个数字——而在于理解背景、选择正确的工具，并在真实情境中解读结果。用历年真题针对你所在考试委员会的具体数据集进行练习，你将把这一必考部分变成可靠的得分来源。
Published by TutorHao | Mathematics Revision Series | aleveler.com
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