📚 AS Cambridge Statistics: Key Vocabulary & Memorisation Guide | AS 剑桥统计：词汇术语速记指南
Statistics at the AS level introduces a precise vocabulary that forms the bedrock of all further statistical reasoning. Without a firm command of terms such as 'hypothesis', 'significance level', or 'random variable', examination questions may become needlessly confusing before any calculation begins. This guide arranges the essential terminology by topic, offers clear definitions, and provides visual mnemonics and word-root clues to make memorisation faster and more reliable.
AS 阶段的统计学引入了一套严谨的术语体系，奠定了所有后续统计推理的基础。如果对 'hypothesis'、'significance level'、'random variable' 等术语没有牢固掌握，考试题目在开始计算之前就可能变得毫无必要地混乱。本指南按主题梳理核心术语，提供清晰定义，并通过视觉速记法和词根提示让记忆更快、更牢固。
1. Populations, Samples and Variables | 总体、样本与变量
A population is the complete collection of all individuals or items that are of interest for a statistical study. Think of it as the entire ocean from which you draw your conclusions. The sample is a subset of the population, selected for actual measurement; it is the bucket of water scooped from that ocean. A sampling frame is a physical list of all population members from which the sample is actually drawn, and a sampling unit is a single individual or item on that list. A variable is any characteristic that can vary from one member of the population to another. If the variable takes numerical values that can be counted or measured, it is a quantitative variable; if it describes a quality or category, it is a qualitative variable. Quantitative variables further split into discrete (countable gaps, such as number of students) and continuous (any value on an interval, such as height).
总体 (population) 是统计研究中所有感兴趣的个体或项目的完整集合，可以想象成你抽取结论的整片海洋。样本 (sample) 是从总体中选出用于实际测量的子集，就像从海洋中舀出一桶水。抽样框 (sampling frame) 是包含所有总体成员的物理列表，样本实际从中抽取；而 抽样单元 (sampling unit) 是该列表中的单个个体。变量 (variable) 是能从总体一个成员变化到另一个成员的任何特征。如果变量取可计数或测量的数值，就是 定量变量 (quantitative variable)；若描述性质或类别，则是 定性变量 (qualitative variable)。定量变量又分为 离散型 (discrete)（存在可数的间隔，如学生人数）和 连续型 (continuous)（可在区间内取任意值，如身高）。
2. Types of Data and Measurement Scales | 数据类型与测量尺度
Primary data is collected firsthand by the researcher for the specific purpose at hand, offering full control but higher cost and time. Secondary data already exists, gathered by someone else for a different purpose, making it quick to obtain though potentially less tailored. Beyond this origin distinction, data sits on a measurement scale. Nominal data uses names or labels with no natural order (eye colour: blue, brown, green). Ordinal data has categories that can be ordered but the gaps between ranks are not meaningful (satisfaction rating: poor, fair, good). Interval data has equal gaps between values but no true zero (temperature in °C). Ratio data has equal intervals and a meaningful zero, allowing statements such as 'twice as much' (mass in kg). Quick mnemonic: the first letters spell NOIR.
一手数据 (primary data) 由研究者为当前特定目的亲自收集，控制力强但成本较高、耗时较长。二手数据 (secondary data) 已经存在，由他人为不同目的收集，获取迅速但可能不够贴合。除了这种来源区别，数据还处于某种测量尺度上。定类 (nominal) 数据使用名称或标签，无自然顺序（如眼睛颜色：蓝、棕、绿）。定序 (ordinal) 数据的类别可排序，但等级之间的差距无实际意义（如满意度评分：差、一般、好）。定距 (interval) 数据有等距间隔但无真正的零点（如摄氏温度）。定比 (ratio) 数据既有等距间隔也有有意义的零点，可以说"是多少倍"（如以千克表示的质量）。速记口诀：四个首字母组成法文单词 NOIR。
3. Sampling Methods and Bias | 抽样方法与偏差
A simple random sample of size n gives every possible sample of that size an equal chance of being chosen, usually achieved with random number generators or lottery methods. Stratified sampling divides the population into distinct groups called strata, then takes a random sample from each, guaranteeing representation of known subgroups. Systematic sampling selects every k-th member from the sampling frame after a random start; it is quick but dangerous if the list contains hidden patterns. Quota sampling is a non-random method where interviewers fill fixed quotas for different categories, often used in market research. Bias is a systematic error that causes the sample to misrepresent the population. Selection bias occurs when some groups are disproportionately excluded; non-response bias arises when those who refuse to participate differ systematically from respondents; measurement bias stems from a faulty instrument or leading questions.
容量为 n 的 简单随机样本 (simple random sample) 使每一个可能样本被选中的概率相等，通常借助随机数生成器或抽签实现。分层抽样 (stratified sampling) 先把总体分成不同的组（称为层），再从每层中随机抽样，保证已知子群体的代表性。系统抽样 (systematic sampling) 在随机起点后每隔第 k 个成员选取一个；执行快捷，但若名单存在隐藏模式则很危险。配额抽样 (quota sampling) 是非随机方法，由调查员按不同类别填写固定配额，常用于市场调查。偏差 (bias) 是导致样本不能代表总体的系统性误差。选择偏差发生在某些群体被不成比例地排除时；无回应偏差出现在拒绝参与的人与回应者存在系统性差异时；测量偏差则源自缺陷工具或诱导性问题。
4. Measures of Central Tendency | 集中趋势度量
The mean (x̄ for a sample, μ for a population) is the arithmetic average of all values. It uses every data point, making it sensitive to outliers. The median is the middle value when data are ordered; it splits the dataset into two halves and resists the influence of extreme values. The mode is the most frequently occurring value, useful for categorical data and for spotting bimodality. Imagine a street of houses: the mean is the centre of mass if each house had the same weight, the median is the house where an equal number of houses lie to the left and right, and the mode is the house colour you see most often.
均值 (mean)（样本记作 x̄，总体记作 μ）是所有数值的算术平均。它使用了每一个数据点，因此对异常值敏感。中位数 (median) 是数据排序后的中间值；它将数据集一分为二，能抵抗极端值的影响。众数 (mode) 是出现频率最高的值，对分类型数据以及发现双峰形态很有用。想象一条街道的房屋：均值是假设每栋房屋质量相同时的质心，中位数是左右房屋数目相等的那一栋，而众数则是你看到最多的房屋颜色。
5. Measures of Dispersion and Position | 离散与位置度量
Range is simply the difference between the maximum and minimum values; it is easy to compute but depends entirely on the two most extreme observations. Interquartile range (IQR) equals the third quartile minus the first quartile (Q₃ − Q₁); it captures the spread of the middle 50% and is robust against outliers. Variance x − μ)² ÷ N; for a sample, the denominator becomes n − 1 to give an unbiased estimate. The standard deviation is the square root of variance, returning the spread measure to the original units. A percentile indicates the value below which a given percentage of observations fall; the 90th percentile, for example, leaves 10% of values above it.
极差 (range) 就是最大值与最小值之差；计算简单，但完全依赖两个最极端的观测值。四分位距 (interquartile range, IQR) 等于上四分位数减下四分位数（Q₃ − Q₁）；它捕捉中间 50% 数据的散布情况，对异常值稳健。方差 (variance) 衡量与均值的平均平方偏差。总体方差 σ² = Σ(x − μ)² ÷ N；样本方差分母变为 n − 1 以给出无偏估计。标准差 (standard deviation) 是方差的平方根，将散布度量还原到原始单位。百分位数 (percentile) 表示给定百分比观测值落入其下的数值；例如第 90 百分位数意味着 10% 的值在其上。
6. Probability Basics and Notation | 概率基础与符号
Probability P(A) measures the likelihood of event A, always between 0 (impossible) and 1 (certain). An experiment is any repeatable process that yields observations; a trial is a single performance of the experiment. The sample space (often denoted S or Ω) is the set of all possible outcomes. An event is a subset of the sample space. Two events are mutually exclusive if they cannot occur simultaneously: P(A ∩ B) = 0. The complement of A, written A' or Aᶜ, consists of all outcomes not in A, with P(A') = 1 − P(A). Two events are independent if the occurrence of one does not affect the probability of the other: P(A ∩ B) = P(A) × P(B).
概率 P(A) 度量事件 A 发生的可能性，始终介于 0（不可能）与 1（必然）之间。试验 (experiment) 是任何可重复产生观测的过程；一次尝试 (trial) 是试验的单次执行。样本空间 (sample space)（常记作 S 或 Ω）是所有可能结果的集合。事件 (event) 是样本空间的子集。若两个事件不能同时发生，则它们 互斥 (mutually exclusive)：P(A ∩ B) = 0。A 的补事件 (complement) 记作 A' 或 Aᶜ，包含所有不在 A 中的结果，且 P(A') = 1 − P(A)。若一个事件的发生不影响另一个事件的概率，则两事件独立 (independent)：P(A ∩ B) = P(A) × P(B)。
7. Conditional Probability and Tree Diagrams | 条件概率与树状图
Conditional probability P(A | hat event A occurs given that event B has already occurred. The fundamental formula is P(A | B) = P(A ∩ h labelled with its conditional probability. The probability of any complete path is the product of the probabilities along its branches. To find the probability of an event that spans multiple paths, sum the probabilities of those paths. Remember: along a branch you multiply; across different branches you add.
条件概率 (conditional probability) P(A | B) 是在事件 B 已经发生的条件下，事件 A 发生的概率。基本公式为 P(A | B) = P(A ∩ B) ÷ P(B)，前提是 P(B) > 0。树状图是处理多阶段试验的强大可视化工具。从节点出发的枝条表示可能的结果，每条枝条标注其条件概率。任意一条完整路径的概率是沿途枝条概率的乘积。若要找出横跨多条路径的事件的概率，只需将各路径概率相加。记住：沿枝条相乘，跨枝条相加。
8. Random Variables and Distributions | 随机变量及其分布
A random variable is a variable whose value is a numerical outcome of a random phenomenon, conventionally denoted by an uppercase letter such as X. A probability distribution lists all possible values of the random variable together with their associated probabilities. For a discrete random variable, the sum of all probabilities must equal 1. The expected value the long-run average if the experiment were repeated indefinitely: E(X) = Σ xᵢ P(X = xᵢ). Variance Var(X) = E[(X − μ)²] = E(X²) − [E(X)]². A probability mass function gives P(X = x) for discrete variables; a probability density function applies to continuous variables, where only intervals have non-zero probabilities. Word-root tip: 'stochastic' comes from Greek stokhos ('aim, guess'), the same root as 'stochastic' processes in statistics.
随机变量 (random variable) 是取值由随机现象决定的变量，习惯上用大写字母（如 X）表示。概率分布 (probability distribution) 列出随机变量的所有可能值及其关联概率。对离散型随机变量 (discrete random variable)，所有概率之和必须等于 1。期望值 (expected value) E(X) 是试验无限重复下的长期平均值：E(X) = Σ xᵢ P(X = xᵢ)。方差 (variance) Var(X) = E[(X − μ)²] = E(X²) − [E(X)]²。概率质量函数 (probability mass function) 在离散变量时给出 P(X = x)；概率密度函数 (probability density function) 适用于连续变量，此时只有区间具有非零概率。词根提示：'stochastic' 源自希腊语 stokhos（目标、猜测），统计学中'随机过程'与此同源。
9. The Binomial Distribution | 二项分布
The binomial distribution models the number of successes in a fixed number of independent trials, each with the same probability of success p. The shorthand is X ~ B(n, p), where n is the number of trials. The conditions are memory-friendly with the acronym FITS: Fixed number of trials, Independent trials, Two possible outcomes (success/failure), and S Same probability p for each trial. The probability of exactly r successes is P(X = r) = ⁿCᵣ × pʳ × (1−p)ⁿ⁻ʳ memory hook: picture a pinball machine with n fixed balls, each hitting a peg and falling 'success' or 'failure' with constant probability; the bell-shaped histogram builds as n grows.
二项分布描述在固定次数独立试验中成功次数的模型，每次试验成功的概率同为 p。简记为 X ~ B(n, p)，其中 n 为试验次数。适用条件可用首字母缩略词 FITS 帮助记忆：F 固定试验次数 (Fixed)、I 试验相互独立 (Independent)、T 两种可能结果 (Two outcomes)、S 每次试验相同成功概率 (Same p)。恰好 r 次成功的概率为 P(X = r) = ⁿCᵣ × pʳ × (1−p)ⁿ⁻ʳ。期望值 E(X) = np，方差 Var(X) = np(1−p)。视觉记忆法：想象一台弹珠机有 n 个固定小球，每个撞击木桩后以恒定概率落入"成功"或"失败"槽；随着 n 增大，直方图逐渐呈现钟形。
10. The Normal Distribution | 正态分布
shaped distribution defined by two parameters: mean μ and standard deviation σ. We write X ~ N(μ, σ²). The total area under the curve equals 1. About 68% of values lie within μ ± σ, roughly 95% 3σ. To standardise, subtract the mean and divide by the standard deviation, obtaining the standard normal Z ~ disation lets us use a single table of probabilities for any normal distribution. The term 'normal' is historically misleading; Gauss called it the 'law of errors' , reflecting its origins in astronomical measurement discrepancies, not because other distributions are 'abnormal'.
正态分布是一个连续、对称的钟形分布，由两个参数定义：均值 μ 和标准差 σ。记作 X ~ N(μ, σ²)。曲线下方面积总和为 1。约 68% 的数值落在 μ ± σ 内，约 95% 在 μ ± 2σ 内，几乎全部落在 μ ± 3σ 内。进行标准化时，减去均值再除以标准差，得到标准正态分布 Z ~ N(0, 1)：Z = (X − μ) ÷ σ。标准化使我们能对任何正态分布使用同一张概率表。'正态'这一术语在历史上具有误导性；高斯称它为'误差定律'，反映其源于天文测量偏差的背景，并非表示其他分布'不正常'。
11. Hypothesis Testing Vocabulary | 假设检验术语
Hypothesis testing is a formal decision-making procedure. The null hypothesis H₀ is the default position, usually stating 'no effect' or 'no difference'. The alternative hypothesis H₁ is the claim we hope to support with evidence. A test statistic is a value calculated from the sample data, used to decide between H₀ and H₁. The significance level α is the probability of rejecting H₀ when it is in fact true — a Type I error. The p-value is the probability, assuming H₀ is true, of obtaining a test statistic at least as extreme as the one observed. If p < α, we reject H₀ and say the result is statistically significant. The critical region is the set of values of the test statistic that lead to rejecting H₀ at the chosen α. A Type II error occurs when H₀ is false but we fail to reject it. Memory image: think of a courtroom where H₀ is 'innocent', α is the acceptable risk of convicting an innocent person, and the p-value is the strength of the evidence the prosecution presents.
假设检验是一套正式的决策程序。零假设 (null hypothesis) H₀ 是默认立场，通常陈述为"无效应"或"无差异"。备择假设 (alternative hypothesis) H₁ 是我们希望用证据支持的主张。检验统计量 (test statistic) 是从样本数据算得的值，用于在 H₀ 和 H₁ 之间作出判断。显著性水平 (significance level) α 是 H₀ 实际为真时我们却拒绝它的概率——这是第一类错误。p 值 (p-value) 是假设 H₀ 为真的条件下，获得至少与观测值一样极端的检验统计量的概率。若 p < α，我们拒绝 H₀，并称结果具有统计显著性。拒绝域 (critical region) 是在所选 α 下导致拒绝 H₀ 的检验统计量取值的集合。当 H₀ 为假但我们未能拒绝它时，发生第二类错误 (Type II error)。记忆意象：想象法庭上 H₀ 是'无罪'，α 是可接受的冤枉无辜者的风险，p 值则是控方呈堂证据的强度。
12. Correlation, Regression and Residuals | 相关、回归与残差
Correlation measures the strength and direction of a linear relationship between two variables. The product moment correlation coefficient (Pearson's r) ranges from −1 (perfect negative) through 0 (none) to +1 (perfect positive). Crucially, correlation does not imply causation. Regression y = a + bx, to model how a response variable y changes with an explanatory variable x. The least squares regression line minimises the sum of squared vertical distances between observed points and the line. A residual is the observed y-value minus the predicted y-value for a given x; residual plots help check whether a linear model is appropriate. An outlier is a point that lies far from the general pattern, while an influential point substantially changes the regression line if removed.
相关 (correlation) 衡量两个变量之间线性关系的强度和方向。积矩相关系数 (product moment correlation coefficient)（Pearson's r）取值范围从 −1（完全负相关）经 0（无线性关系）到 +1（完全正相关）。关键点：相关不蕴含因果关系。回归 (regression) 用一条直线（通常形式为 y = a + bx）来建模响应变量 y 如何随解释变量 x 变化。最小二乘回归线 (least squares regression line) 使观测点与直线之间竖直距离的平方和最小。残差 (residual) 是给定 x 下观测 y 值减去预测 y 值；残差图有助于检验线性模型是否恰当。异常值 (outlier) 是远离总体形态的数据点，而强影响点 (influential point) 则在移除后会显著改变回归线。
Published by TutorHao | AS Statistics 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