📚 AQA International A-level FS2 Statistics: Complete Revision Guide | AQA 国际 A-Level 进阶统计 2 完整复习指南
Welcome to the TutorHao revision guide for AQA International A-level Unit FS2 (Further Statistics 2). This unit extends your understanding of probability, statistical inference, and hypothesis testing. The exam rewards both accurate calculation and clear interpretation.
欢迎阅读 TutorHao 的 AQA 国际 A-level FS2（进阶统计 2）复习指南。本单元扩展了概率、统计推断和假设检验的知识。考试既要求准确计算，也要求清晰的解释。
1. Overview and Assessment Structure | 单元概览与考试结构
FS2 is usually assessed by a written paper lasting 90 minutes or 2 hours, depending on the specification route. Questions mix routine calculations with applied problem solving and interpretation, so you need to show full working and state conclusions in context.
FS2 通常通过 90 分钟或 2 小时的笔试进行评估，具体取决于考试路线。题目混合了常规计算、应用问题求解和解释，因此你需要展示完整步骤，并结合实际背景陈述结论。
Key topics tested include probability generating functions, continuous distributions, estimation, confidence intervals, hypothesis testing, chi-squared tests, and correlation. A solid command of algebraic manipulation and calculator use is essential.
重点考查内容包括概率生成函数、连续分布、估计、置信区间、假设检验、卡方检验和相关分析。扎实的代数运算能力和计算器使用能力至关重要。
- Probability generating functions | 概率生成函数
- Continuous random variables and distributions | 连续随机变量与分布
- Estimation and confidence intervals | 估计与置信区间
- Hypothesis testing and errors | 假设检验与错误
- Chi-squared tests | 卡方检验
- Correlation and regression | 相关与回归
2. Probability Generating Functions | 概率生成函数
For a discrete random variable X taking non-negative integer values, the probability generating function is GX(t) = E(tX) = Σ pₓ tˣ, where pₓ = P(X = x). The PGF is defined for |t| ≤ 1 and always satisfies GX(1) = 1.
对于取非负整数值的离散随机变量 X，概率生成函数为 GX(t) = E(tX) = Σ pₓ tˣ，其中 pₓ = P(X = x)。PGF 在 |t| ≤ 1 时定义，并且始终满足 GX(1) = 1。
PGFs are useful because derivatives give moments: G'_X(1) = E(X) and G''_X(1) = E(X(X − 1)). Therefore Var(X) = G''_X(1) + G'_X(1) − [G'_X(1)]². This avoids direct summation in many cases.
PGF 的用处在于其导数可以给出矩：G'_X(1) = E(X)，G''_X(1) = E(X(X − 1))。因此 Var(X) = G''_X(1) + G'_X(1) − [G'_X(1)]²。在许多情况下，这可以避免直接求和。
If X and Y are independent, then GX+Y(t) = GX(t)GY(t). Recognising standard PGFs also helps: for X ~ Binomial(n, p), GX(t) = (q + pt)ⁿ; for X ~ Poisson(λ), GX(t) = e^(λ(t − 1)).
如果 X 和 Y 独立，则 GX+Y(t) = GX(t)GY(t)。识别标准 PGF 也有帮助：若 X ~ 二项分布 Binomial(n, p)，则 GX(t) = (q + pt)ⁿ；若 X ~ 泊松分布 Poisson(λ)，则 GX(t) = e^(λ(t − 1))。
3. Continuous Random Variables and Key Distributions | 连续随机变量与重要分布
The exponential distribution with parameter λ has probability density function f(x) = λe−λx mean is 1/λ and its variance is 1/λ². The cumulative distribution function is F(x) = 1 − e−λx.
参数为 λ 的指数分布的概率密度函数为 f(x) = λe−λx，其中 x ≥ 0。其均值为 1/λ，方差为 1/λ²。累积分布函数为 F(x) = 1 − e−λx。
For integer k, the gamma distribution Gamma(k, λ) has density f(x) = λᵏ xk−1 e−λx / (k−1)!, with mean k/λ and variance k/λ². The sum of k independent exponential(λ) variables follows Gamma(k, λ), linking the two distributions.
对于整数 k，伽马分布 Gamma(k, λ) 的密度函数为 f(x) = λᵏ xk−1 e−λx / (k−1)!，均值为 k/λ，方差为 k/λ²。k 个独立的指数分布 exponential(λ) 变量之和服从 Gamma(k, λ)，这建立了两个分布之间的联系。
rval: P(a < X < b) = ∫ₐᵇ f(x) dx. Always check that the total area under the density is 1.
要求连续随机变量的概率，需要对密度函数在所需区间上积分：P(a < X < b) = ∫ₐᵇ f(x) dx。务必检查密度函数下的总面积为 1。
4. Sums and Linear Combinations of Random Variables | 随机变量的和与线性组合
+ b²Var(Y). This rule also applies to Var(X − Y) because variances add under independence.
对于任意随机变量 X 和 Y，期望是线性的：E(aX + bY) = aE(X) + bE(Y)。如果 X 和 Y 独立，则 Var(aX + bY) = a²Var(X) + b²Var(Y)。该规则也适用于 Var(X − Y)，因为在独立条件下方差相加。
For a sample mean X̄ independent and identically distributed variables with mean μ and variance σ², we have E(X̄) = μ and Var(X̄ decreasing variance is the mathematical reason larger samples give more precise estimates.
对于来自 n 个独立同分布变量、均值为 μ、方差为 σ² 的样本均值 X̄，我们有 E(X̄) = μ，Var(X̄) = σ²/n。方差随样本量增大而减小，这正是大样本估计更精确的数学原因。
2, independence is usually stated, but be ready to use this general formula in bivariate questions.
当变量不独立时，需要加入协方差项：Var(X + Y) = Var(X) + Var(Y) + 2Cov(X, Y)。在 FS2 中通常会明确给出独立性，但在双变量问题中要会使用这个一般公式。
5. Point Estimation and Unbiased Estimators | 点估计与无偏估计量
An estimator T for an unknown parameter θ is unbiased if E(T) = θ. The sample mean X̄ is an unbiased estimator of the population mean μ because E(X̄) = μ, regardless of the underlying distribution.
如果估计量 T 对未知参数 θ 满足 E(T) = θ，则称 T 为无偏估计量。样本均值 X̄ 是总体均值 μ 的无偏估计量，因为 E(X̄) = μ，无论总体分布如何。
The sample variance S² = Σ(Xᵢ − X̄ divide by n instead of n − 1, the estimator is biased, with expected value E[(1/n)Σ(Xᵢ − X̄)²] = (n − 1)σ²/n.
样本方差 S² = Σ(Xᵢ − X̄)² / (n − 1) 是 σ² 的无偏估计量。如果除以 n 而不是 n − 1，该估计量就是有偏的，其期望为 E[(1/n)Σ(Xᵢ − X̄)²] = (n − 1)σ²/n。
The standard error of the sample mean is σ/√n. When σ is unknown, use the estimated standard error s/√n, where s is the sample standard deviation. This distinction is important for confidence intervals and hypothesis tests.
样本均值的标准误为 σ/√n。当 σ 未知时，使用估计标准误 s/√n，其中 s 为样本标准差。这一区别对置信区间和假设检验非常重要。
6. Confidence Intervals | 置信区间
A 95% confidence interval for the population mean μ, when σ is known, is x̄ ± 1.96 × σ/√n. The multiplier comes from the standard normal distribution: 1.645 for 90%, 1.96 for 95%, and 2.576 for 99% confidence.
当 σ 已知时，总体均值 μ 的 95% 置信区间为 x̄ ± 1.96 × σ/√n。乘数来自标准正态分布：90% 置信水平为 1.645，95% 为 1.96，99% 为 2.576。
nown and the population is normal, use the t-distribution with n − 1 degrees of freedom: x̄ ± t × s/√n. The t multiplier is larger than the corresponding z multiplier, reflecting the extra uncertainty from estimating σ.
如果 σ 未知且总体服从正态分布，则使用自由度为 n − 1 的 t 分布：x̄ ± t × s/√n。t 乘数比相应的 z 乘数更大，反映了估计 σ 带来的额外不确定性。
A correct interpretation is: if we repeated the sampling process many times, about 95% of the intervals constructed would contain μ. We do not say that μ has a 95% probability of lying in one particular interval.
正确的解释是：如果我们多次重复抽样过程，大约 95% 的构造区间会包含 μ。我们不能说 μ 在一个特定区间内的概率为 95%。
7. Hypothesis Testing: Errors, Power, and p-values | 假设检验：错误、功效与 p 值
A hypothesis test starts with a null hypothesis H0 and an alternative hypothesis H1. The test statistic is compared with a critical value, or the p-value is compared with the significance level α. A small p-value gives evidence against H0.
假设检验从原假设 H0 和备择假设 H1 开始。将检验统计量与临界值比较，或将 p 值与显著性水平 α 比较。p 值很小表示有证据反对 H0。
A Type I error occurs when H0 is true but we reject it; its probability is α, the significance level. A Type II error occurs when H1 is true but we fail to reject H0; its probability is denoted β. The power of a test is 1 − β, the probability of correctly rejecting a false H0.
第一类错误发生在 H0 为真但被拒绝时；其概率为 α，即显著性水平。第二类错误发生在 H1 为真但未能拒绝 H0 时；其概率记为 β。检验的功效为 1 − β，即正确拒绝错误 H0 的概率。
Power increases when the sample size increases, when the true effect size is larger, or when α is increased. Exam questions often ask you to calculate β or power for a specified critical region.
当样本量增大、真实效应量更大或 α 增大时，功效会增加。考试题经常要求针对特定拒绝域计算 β 或功效。
8. Chi-Squared Goodness-of-Fit Tests | 卡方拟合优度检验
Σ (O − E)² / E, where O are observed frequencies and E are expected frequencies under H0.
卡方拟合优度检验用于检验观察频数是否符合指定分布。检验统计量为 X² = Σ (O − E)² / E，其中 O 为观察频数，E 为 H0 下的期望频数。
mated parameter. For example, testing a Poisson mean estimated from data gives df = k − 2.
自由度通常为（类别数 − 1），但如果从数据中估计参数，则每估计一个参数再减去一个自由度。例如，检验由数据估计均值的泊松分布时，df = k − 2。
cted frequencies are too small, combine adjacent categories where sensible before applying the test.
该近似的有效性要求期望频数足够大，通常所有 E ≥ 5。如果某些期望频数过小，应在合理的情况下合并相邻类别后再进行检验。
9. Chi-Squared Tests for Contingency Tables | 列联表卡方检验
A contingency table tests whether two categorical variables are independent. The expected frequency in each cell is Eij w total × column total) / grand total. The test statistic is the same X² = Σ (O − E)² / E.
列联表用于检验两个分类变量是否独立。每个单元格的期望频数为 Eij =（行合计 × 列合计）/ 总计。检验统计量同样是 X² = Σ (O − E)² / E。
), where r is the number of rows and c is the number of columns. The null hypothesis states that the two variables are independent.
自由度为 df = (r − 1)(c − 1)，其中 r 为行数，c 为列数。原假设表明两个变量相互独立。
All expected frequencies should be at least 5 for the chi-squared approximation to be reliable. If the test statistic exceeds the critical value, there is significant evidence of an association between the variables.
所有期望频数应至少为 5，卡方近似才可靠。如果检验统计量超过临界值，则表明有显著证据证明变量之间存在关联。
10. Correlation, Regression, and Exam Strategy | 相关、回归与考试策略
The product moment correlation coefficient is r = Sxy / √Sxx Syy, where Sxx = Σ(x − x̄)², Syy = Σ(y − ȳ)², and Sxy = Σ(x − x̄)(y − ȳ). It measures the strength and direction of a linear relationship from −1 to 1.
积矩相关系数为 r = Sxy / √Sxx Syy，其中 Sxx = Σ(x − x̄)²，Syy = Σ(y − ȳ)²，Sxy = Σ(x − x̄)(y − ȳ)。它衡量线性关系的强度和方向，取值范围为 −1 到 1。
To test H0: ρ = 0, use t = r√n − 2 / √1 − r² with n − 2 degrees of freedom. Spearman's rank correlation can be used if the data are not linear or contain outliers.
检验 H0: ρ = 0 时，使用 t = r√n − 2 / √1 − r²，自由度为 n − 2。如果数据不是线性或包含离群值，可以使用 Spearman 秩相关。
For linear regression, the least squares line is y = a + bx, where b = Sxy / Sxx and a = ȳ − b x̄ redicted y-values; a plot of residuals can reveal whether the linear model is appropriate.
对于线性回归，最小二乘直线为 y = a + bx，其中 b = Sxy / Sxx，a = ȳ − b x̄。残差是观察 y 值与预测 y 值之差；残差图可以揭示线性模型是否合适。
In the exam, write down hypotheses with correct notation, show distributions of test statistics, state degrees of freedom, and compare critical values or p-values clearly. Always finish with a conclusion in the context of the original problem, not just a mathematical statement.
在考试中，要写出正确符号的假设，给出检验统计量的分布，说明自由度，并清晰地比较临界值或 p 值。最后一定要结合原始问题的背景给出结论，而不只是数学陈述。
Published by TutorHao | Further Statistics 2 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