📚 A-Level Edexcel Maths: Edexcel AS and A Level Further Mathematics - Further Statistics 2 Question Analysis | A-Level Edexcel 数学：Further Statistics 2 题型解析
Further Statistics 2 (FS2) builds on the foundations of AS Further Statistics and stretches students into advanced inferential methods, including a wide range of hypothesis tests, estimation theory, and stochastic processes. This article breaks down the typical question types found in the Edexcel FS2 examination and highlights key techniques for tackling them effectively. By understanding the structure of each topic and the common pitfalls, candidates can approach their revision with clarity and confidence.
Further Statistics 2（FS2）在 AS 高等统计学的基础上进一步拓展，涵盖了一系列假设检验方法、估计理论以及随机过程。本文详细解析 Edexcel FS2 考试中的典型题型，并突出高效解题的关键技巧。理解了每个主题的结构和常见陷阱后，考生就可以清晰、自信地开展复习。
1. Hypothesis Testing Fundamentals | 假设检验基础
Every FS2 exam question on hypothesis testing expects you to define hypotheses, choose an appropriate test, compute a test statistic, and draw a conclusion in context. The null hypothesis H₀ usually represents no effect or no difference, while the alternative H₁ states what we suspect. You must always state whether the test is one‑tailed or two‑tailed and justify the choice based on the wording of the problem.
FS2 考试中每一道假设检验题都要求你建立假设、选择合适的检验、计算检验统计量并在实际情境中得出结论。原假设 H₀ 通常表示无效应或无差异，备择假设 H₁ 则陈述我们所怀疑的方向。你必须始终说明检验是单尾还是双尾，并根据题目文字说明理由。
- Typical question: 'Test, at the 5% significance level, whether the new drug reduces recovery time.' 典型题目：“在 5% 显著性水平下检验新药是否缩短康复时间。”
- Key points: State H₀: μ = μ₀, H₁: μ < μ₀ (one‑tailed). Compute test statistic and compare with critical value or use p‑value. Write a conclusion referring to evidence against H₀. 关键点：写出 H₀: μ = μ₀, H₁: μ < μ₀ (单尾)。计算检验统计量并与临界值比较，或使用 p 值。写出结论，说明是否有证据拒绝 H₀。
2. t‑test for a Single Mean | 单样本均值的 t 检验
When the population variance σ² is unknown and you have a small sample, the t‑test is the standard method. The test statistic t = (x̄ − μ₀) / (s / √n) follows a t‑distribution with n − 1 degrees of freedom. Questions often provide summary statistics and ask you to test a claim about the population mean.
当总体方差 σ² 未知且样本量较小时，t 检验是标准方法。检验统计量 t = (x̄ − μ₀) / (s / √n) 服从自由度为 n − 1 的 t 分布。题目通常给出汇总统计量，要求检验关于总体均值的声称。
Typical exam tasks include constructing a confidence interval for μ using the t‑distribution and performing a two‑sample t‑test when variances are assumed equal (pooled estimate). Always check whether the question assumes σ² unknown and the data are normally distributed.
典型考试任务包括使用 t 分布构造 μ 的置信区间，以及在假定方差相等时进行两样本 t 检验（合并方差估计）。务必检查题目是否假定 σ² 未知以及数据是否来自正态分布。
3. Chi‑Squared Goodness‑of‑Fit Test | 卡方拟合优度检验
The chi‑squared goodness‑of‑ m are usually (number of categories − 1 − number of estimated parameters). A common mistake is forgetting to subtract parameters estimated from the data, such as the mean or total probability.
卡方拟合优度检验用于判断观测频数是否与理论分布吻合。检验统计量为 χ² = Σ ((Oᵢ − Eᵢ)² / Eᵢ)。自由度通常为（类别数 − 1 − 估计参数的个数）。常见错误是忘记减去从数据中估计的参数，例如均值或总概率。
| Step | 注意 |
|---|---|
| Group data into k classes with non‑zero expected frequencies | 确保所有期望频数 ≥ 5（或合并类别） |
| Write hypotheses H₀: data follow a specified distribution; H₁: they do not | 清楚写出分布名称 |
| Calculate χ² and compare with χ²_critical from tables | 使用正确的自由度 |
4. Chi‑Squared Test for Independence | 卡方独立性检验
In a two‑ frequencies are calculated under the assumption of independence: Eᵢ = (row total × column total) / grand total. Degrees of freedom: (rows − 1)(columns − 1).
在双向列联表中，我们检验两个类别变量是否有关联。检验统计量同样为 χ² = Σ ((Oᵢ − Eᵢ)² / Eᵢ)，期望频数是在独立性假设下计算的：Eᵢ = (行合计 × 列合计) / 总计。自由度为 (行数 − 1) × (列数 − 1)。
Questions often require interpreting a small expected frequency warning: if any Eᵢ < 5, Fisher’s exact test may be needed (though FS2 usually avoids this). Instead, the paper will either instruct you to combine categories or remind you that the test is still valid provided no more than 20% of Eᵢ are below 5.
题目常要求解释期望频数过小的警告：如果某个 Eᵢ < 5，可能需要使用 Fisher 精确检验（不过 FS2 通常避免这种情况）。考卷要么指示你合并类别，要么提醒只要低于 5 的期望频数占比不超过 20%，检验仍然有效。
5. F‑test for Equality of Variances | 方差齐性的 F 检验
Before conducting a two‑sample t‑test, we often check whether the population variances are equal using an F‑ istic is F = s₁² / s₂² (with s₁² > s₂² to give F > 1). Under H₀: σ₁² = σ₂², F follows an F‑distribution with (n₁ − 1, n₂ − 1) degrees of freedom.
在进行两样本 t 检验之前，我们通常使用 F 检验检查总体方差是否相等。检验统计量为 F = s₁² / s₂²（将较大的样本方差放在分子以得到 F > 1）。在 H₀: σ₁² = σ₂² 下，F 服从自由度为 (n₁ − 1, n₂ − 1) 的 F 分布。
In the exam, you will be given a significance level and need to find the appropriate critical value from F‑tables. This is usually a two‑tailed test, but tables are often one‑tailed; you must halve the significance level when looking up the upper critical value.
考试中会给出显著性水平，你需要从 F 分布表中查找适当的临界值。这通常是双尾检验，但表格常提供单尾临界值；查阅上侧临界值时需要将显著性水平减半。
6. Maximum Likelihood Estimation (MLE) | 最大似然估计
Maximum likelihood estimation is a powerful technique for finding point estimates of parameters. For independent observations from a probability distribution, the likelihood function L(θ) is the product of the probability density (or mass) functions. The MLE is the value θ̂ that maximises L(θ), usually found by setting the derivative of the log‑likelihood to zero.
最大似然估计是寻找参数点估计的强大方法。对于来自某个概率分布的独立观测值，似然函数 L(θ) 是各概率密度（或质量）函数的乘积。最大似然估计 θ̂ 是使 L(θ) 最大化的值，通常通过将对数似然函数的导数置零来求得。
ℓ(θ) = ln L(θ) = Σ ln f(xᵢ | θ), → dℓ/dθ = 0 solve for θ̂
Typical FS2 questions involve finding MLEs for parameters of the Poisson, binomial, exponential or geometric distributions. You may also need to verify that a critical point is indeed a maximum (second derivative negative) or state the approximate distribution of the MLE for large samples (asymptotic normality).
FS2 典型题目涉及求出泊松、二项、指数或几何分布参数的 MLE。你还可能需要验证临界点确实是最大值（二阶导数为负），或者陈述大样本下 MLE 的近似分布（渐近正态性）。
7. Linear Regression and Hypothesis Tests for Correlation | 线性回归与相关系数的假设检验
ng formal hypothesis tests for the population correlation coefficient ρ. Using the product‑moment correlation coefficient r, we test H₀: ρ = 0 against H₁: ρ ≠ 0 (or one‑tailed) with a t‑test or by comparing r to critical values from tables.
FS2 通过引入总体相关系数 ρ 的正式假设检验，加深了对线性回归的理解。使用积差相关系数 r，通过 t 检验或将 r 与查表临界值比较，来检验 H₀: ρ = 0 对 H₁: ρ ≠ 0（或单尾）。
appropriate. You must know how to calculate rₛ and test H₀: ρₛ = 0 using critical values.
当数据不符合正态分布或秩次更合适时，还会考察 Spearman 秩相关系数。你必须掌握如何计算 rₛ 并使用临界值检验 H₀: ρₛ = 0。
8. Markov Chains | 马尔可夫链
Two‑ ite down P, find the probability distribution after n steps using Pⁿ, and compute the long‑run steady‑state distribution π such that πP = π with Σ πᵢ = 1.
两状态马尔可夫链频繁出现。转移矩阵 P 给出从一种状态转移到另一种状态的概率。关键任务是：写出 P，利用 Pⁿ 求出 n 步后的概率分布，并计算长期稳态分布 π，满足 πP = π 且 Σ πᵢ = 1。
Typical questions may ask: 'Given that today is sunny, find the probability it is sunny in three days´ time.' itial vector v = [1, 0] and compute vP³. Another common request is to solve for the steady‑state directly from the equations π₁ = p₁₁π₁ + p₂₁π₂, π₂ = p₁₂π₁ + p₂₂π₂ and π₁ + π₂ = 1.
典型题目可能问：“已知今天是晴天，求三天后是晴天的概率。”你需要建立初始向量 v = [1, 0] 并计算 vP³。另一个常见要求是直接通过方程组 π₁ = p₁₁π₁ + p₂₁π₂, π₂ = p₁₂π₁ + p₂₂π₂ 以及 π₁ + π₂ = 1 求解稳态。
9. Confidence Intervals for Variance and Ratio of Variances | 方差和方差比的置信区间
While FS1 focuses on confidence intervals for means, FS2 extends this to the population variance σ². Using the fact that (n−1)s²/σ² ~ χ²(n−1), you can derive a (1 − α) confidence interval for σ²:
FS1 侧重于均值的置信区间，而 FS2 将其扩展到总体方差 σ²。利用 (n−1)s²/σ² ~ χ²(n−1) 这一事实，可以推出 σ² 的 (1 − α) 置信区间：
( (n−1)s² / χ²α/2, n-1 , (n−1)s² / χ²1−α/2, n-1 )
ilarly, for the ratio of two variances σ₁²/σ₂², an interval can be constructed using the F‑distribution: (s₁²/s₂²) × (1 / Fα/2, ν₂, ν₁) and (s₁²/s₂²) × Fα/2, ν₁, ν₂, where the degrees of freedom are those of the corresponding samples.
同样地，对于两个方差的比值 σ₁²/σ₂²，可以利用 F 分布构造区间：(s₁²/s₂²) × (1 / Fα/2, ν₂, ν₁) 和 (s₁²/s₂²) × Fα/2, ν₁, ν₂，其中自由度为相应样本的自由度。
10. Quality Control Charts: Shewhart and CUSUM | 质量控制图：休哈特图与累积和图
rts are used to monitor a production process. A Shewhart chart for the sample mean plots x̄ against sample number, with action lines at μ ± 3σ/√n and warning lines at μ ± 2σ/√n. Decisions are based on whether points lie outside the action limits or whether a run of points suggests the process is out of control.
控制图用于监控生产过程。样本均值的休哈特图将 x̄ 相对于样本序号作图，行动线位于 μ ± 3σ/√n，警戒线位于 μ ± 2σ/√n。决策依据有点是否落在行动界外，或者是否存在连续点表明过程失控。
harts are more sensitive to small shifts in the process mean. The cumulative sum Sₙ = Σ (x̄ USUM exceeds a predetermined decision interval. The mask method is often tested, where a V‑shaped mask is placed on the chart to detect significant changes.
CUSUM（累积和）图对过程均值的微小偏移更为敏感。将累积和 Sₙ = Σ (x̄ᵢ − μ₀) 作图，当累积和超过预定的判定区间时做出决定。常考察掩模法，即在图上放置 V 形掩模以检测显著变化。
11. Analysis of Variance (ANOVA) – One‑way Classification | 方差分析 – 单因素分类
One‑ the means of several populations. The total sum of squares SST is partitioned into between‑groups sum of squares SSB and within‑groups (error) sum of squares SSW. The F‑statistic = (SSB/(k−1)) / (SSW/(N−k)) is compared with F_(k−1, N−k, α).
单因素方差分析用于检验多个总体均值之间的差异。总平方和 SST 分解为组间平方和 SSB 和组内（误差）平方和 SSW。F 统计量 = (SSB/(k−1)) / (SSW/(N−k)) 与 F_(k−1, N−k, α) 比较。
| Source | SS | df | MS | F |
| Between | Σ nᵢ (x̄ᵢ − x̄)² | k−1 | MSB = SSB/(k−1) | MSB/MSW |
| Within | Σ Σ (xᵢⱼ − x̄ᵢ)² | N−k | MSW = SSW/(N−k) |
tes whether there is sufficient evidence to reject H₀: μ₁ = μ₂ = ... = μₖ. A significant F‑test suggests not all means are equal; follow‑up tests (e.g., Tukey) may be requested in context.
结论说明是否有足够证据拒绝 H₀: μ₁ = μ₂ = ... = μₖ。显著的 F 检验表明并非所有均值相等；题目可能要求根据情境进行后续检验（如 Tukey）。
12. Errors, Power and Sample Size Determination | 错误、检验功效与样本量确定
β = P(Type II error). FS2 questions often ask you to calculate power for a given alternative or to find the sample size required to achieve a specified power.
理解第一类错误（H₀ 为真却拒绝）和第二类错误（H₀ 为假却未拒绝）至关重要。检验功效为 1 − β，其中 β = P(第二类错误)。FS2 题目经常要求计算指定备择假设下的功效，或求达到特定功效所需的样本量。
In the context of t‑tests, these calculations involve the non‑central t‑distribution. You may be given operating characteristic curves or a normal approximation. The reasoning is: specify the effect size you wish to detect, choose significance level α, then solve for n such that β is small enough.
在 t 检验的情境下，这些计算涉及非中心 t 分布。题目可能给出操作特性曲线或正态近似。推理步骤是：明确想要检测的效应大小，选择显著性水平 α，然后求解 n 使得 β 足够小。
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