📚 Common Mistakes in Edexcel Further Statistics 2 | 爱德思进阶统计学2易错点总结
Edexcel Further Statistics 2 demands precision across probability generating functions, hypothesis testing, regression, and estimation. Even strong candidates lose marks by overlooking convergence conditions, mixing up degrees of freedom, or misapplying formula sheets under pressure. This article highlights the most persistent pitfalls and shows how to avoid them, helping you turn avoidable errors into secure marks.
Edexcel 进阶统计学 2 涵盖概率生成函数、假设检验、回归与估计，对细节要求极高。即使是基础扎实的学生，也常因忽略收敛条件、混淆自由度或在紧张时误用公式而失分。本文将梳理最顽固的易错点，并给出避错方法，助你把可避免的错误转化为稳拿的分数。
1. PGF Domain & Convergence | 概率生成函数定义域与收敛性
A probability generating function GX(t) = E(tX) is only guaranteed for |t| ≤ 1. Many candidates write GX(t) as a power series and then blindly set t = 2 to 'find' g probabilities, producing nonsense. Always verify that the value you substitute lies in the radius of convergence.
概率生成函数 GX(t) = E(tX) 只有在 |t| ≤ 1 时才有保证。许多考生将 GX(t) 写成幂级数后，就盲目代入 t = 2 去“求”概率，得出无意义的结果。务必验证所代的值落在收敛半径内。
When establishing GX s a check, remember that the sum of probabilities must equal 1. Differentiate correctly: G'X(t) involves tx-1, and G'X forgetting to multiply by the power when differentiating a polynomial term or evaluating G'X(1) too early before simplifying.
用 GX(1) = 1 进行验证时，记住概率总和必须为 1。要正确求导：G'X(t) 涉及 tx-1，且 G'X(1) = E(X)。常见失误是在对多项式项求导时忘记乘以幂次，或在未化简的情况下过早计算 G'X(1)。
For a sum of independent random variables, the PGF of the sum is the product of the individual PGFs. Candidates sometimes apply this rule when the variables are not independent – always check for independence stated in the question.
对于独立随机变量之和，其 PGF 是各变量 PGF 的乘积。考生有时在变量不独立时也使用这条规则——务必检查题目是否声明了独立性。
2. MGF Derivatives & Moment Extraction | 矩生成函数求导与矩提取
The moment generating function MX(t) = E(etX orgetting to substitute t = 0 after differentiation, leaving E(X) expressed in terms of t.
矩生成函数 MX(t) = E(etX) 求导时必须格外小心。要得到第 r 阶原点矩，需要求导 r 次再代入 t = 0。典型错误是求导后忘记将 t = 0 代入，导致 E(X) 的表达式中还含有 t。
When working with linear combinations, canditates often misremember that MaX+b(t) = ebt MX(at). Write the exponent correctly; a single sign error in 'b' can ruin variance calculations. Also, identifying a distribution from its MGF requires recognising standard forms such as the binomial MGF (q + p et)n – don't hastily match without checking the whole structure.
处理线性组合时，考生常记错 MaX+b(t) = ebt MX(at)。要正确书写指数部分；只要 'b' 的符号错了，整个方差计算就会崩溃。另外，从 MGF 识别分布需要认出标准形式，比如二项分布的 MGF (q + p et)n——不要在没有检查整体结构的情况下匆忙匹配。
3. Joint Distributions & Marginal Pitfalls | 联合分布与边缘概率误区
When given a joint probability mass function p(x,y), summing over all y to find the marginal pX(x) seems simple. Yet candidates frequently sum only over a subset or forget to verify that the total probability sums to 1. Always sum carefully: pX(x) = Σy p(x,y).
给定联合概率质量函数 p(x,y) 时，对所有 y 求和以得到边缘分布 pX(x) 看似简单。但考生常常只对部分 y 求和，或忘记验证概率总和为 1。务必仔细求和：pX(x) = Σy p(x,y)。
Independence is tested by checking p(x,y) = pX(x) pY(y) for all x,y. Many lose marks by asserting independence after checking only one pair. Provide a counterexample or verify all cells. Moreover, remember that zero covariance does not imply independence; counterexamples are common in Further Statistics 2.
独立性通过检查所有 x,y 是否满足 p(x,y) = pX(x) pY(y) 来判断。许多人只检验了一对数值就声称独立，因此丢分。要提供反例或逐个验证所有单元格。此外，记住协方差为零并不蕴含独立性；进阶统计学 2 中常有反例。
4. Covariance Computation Errors | 协方差计算错误
rite formula, but calculating E(XY) from a joint table often goes wrong. When using coded data or grouped frequencies, always multiply the x, y, and probability (or frequency) correctly before summing. A single omission of a cross term can cost a whole question.
Cov(X,Y) = E(XY) − E(X)E(Y) 是人见人爱的公式，但从联合表格计算 E(XY) 时经常出错。使用编码或分组频数时，务必先将 x、y 和概率（或频数）正确相乘再求和。漏掉一个交叉项就可能葬送整道题。
When switching between population and sample covariance, note the denominator: population covariance divides by N, while sample covariance typically divides by n−1 (or uses a different formula). Students often treat Σ(x−x̄)(y−ȳ) as equal to Σxy − n x̄ ȳ without adjusting for the correct divisor.
在总体协方差和样本协方差之间切换时，注意分母：总体协方差除以 N，而样本协方差通常除以 n−1（或使用不同公式）。学生常把 Σ(x−x̄)(y−ȳ) 等同为 Σxy − n x̄ ȳ，却未根据正确的除数加以调整。
5. Regression Line Swaps | 回归线方向混淆
on x (y = a + bx) and x on y (x = c + dy) are not simply algebraic rearrangements. A grade‑costing error is using the slope formula b = Sxy/Sxx for x on y regression, or mixing up which variable is the response. Always identify the response variable before writing the equation.
在二元数据中，y 对 x 的回归线 (y = a + bx) 与 x 对 y 的回归线 (x = c + dy) 并非简单的代数变形。一个令人痛心的错误是把 y 对 x 的斜率公式 b = Sxy/Sxx 用在 x 对 y 的回归上，或者混淆了响应变量。务必先确定响应变量再写方程。
Another pitfall is interpreting the intercept outside the data range. The line is only valid for interpolation within the observed x-values; extrapolation predictions are unreliable. Even if the exam asks for an estimate, caveat your answer if it lies far from the data.
另一个陷阱是在数据范围之外解读截距。回归线仅对观测 x 值范围内的内插有效；外推预测不可靠。即使考题要求估算，如果估算值远离数据范围，最好在答案中加以说明。
6. Chi-Squared Degrees of Freedom | 卡方检验自由度迷失
Degrees of freedom (ν) cause persistent trouble. For a goodness‑of‑ y table test, ν = (rows − 1)(columns − 1). Mixing these two formulas is a classic mistake.
自由度 (ν) 经常惹祸。对于拟合优度检验，ν = 类别数 − 1 − 所估计的参数量。考生常忘记减掉从样本中估计的参数。对于列联表检验，ν = (行数 − 1)(列数 − 1)。混淆这两个公式是经典错误。
the test statistic is χ² = Σ (O−E)²/E; use observed and expected counts, not percentages.
期望频数必须全部不小于 5，卡方近似才有效。当某个期望频数小于 5 时，应合并相邻类别直到满足条件。完全忽略这条规则会导致检验失效。同时，检验统计量为 χ² = Σ (O−E)²/E；要使用观测频数和期望频数，而非百分比。
7. Hypothesis Test Power & Errors | 假设检验功效与错误类型
Type I error (rejecting a true null) and Type II error (failing to reject a false null) are frequently confused. A typical slip is labelling a test's significance level as the probability of a type II error – it is actually P(Type I error). Power is 1 − P(Type II error), and calculating it requires a specific alternative hypothesis.
第一类错误（拒绝真的原假设）与第二类错误（未能拒绝假的原假设）常被混淆。典型失误是把检验的显著性水平标为第二类错误的概率——它其实是 P(第一类错误)。功效 = 1 − P(第二类错误)，计算功效需要一个具体的备择假设。
Many questions ask for the power function or power at a point; students often stop after finding the critical region. You must then find the probability that the test statistic falls in the critical region under the alternative distribution. Watch out for whether the test is one‑tailed or two‑tailed when setting up the critical region.
许多题目要求求功效函数或某点处的功效；学生常常在找到拒绝域后就停住了。必须再求出在备择分布下检验统计量落入拒绝域的概率。在建立拒绝域时，注意是单尾还是双尾检验。
8. Confidence Interval Misapplications | 置信区间误用
A confidence interval for a mean is x̄ ± z × (σ/√n) when σ is known, but when σ is unknown (the usual case) you must use the t‑distribution: x̄ ± tn−1 × (s/√n). Forgetting to swap z for t loses marks even if the numbers barely change. Also, check the sample size – a large n does not automatically permit use of z; the correct distribution depends on what is known.
均值的置信区间：当 σ 已知时为 x̄ ± z × (σ/√n)，但当 σ 未知（通常情形）时必须用 t 分布：x̄ ± tn−1 × (s/√n)。忘记把 z 换成 t，即使数值变化不大，也会丢分。此外要检查样本量——大 n 并不自动允许使用 z；正确的分布取决于已知信息。
For a difference of means, the pooled variance formula appears frequently. A common error is using the wrong degrees of freedom for the t‑value, or failing to check the assumption of equal population variances when the pooled estimator is used. In Further Statistics 2, always state any assumptions you are making.
对于均值差，合并方差公式经常出现。常见错误是用了错误的 t 分布自由度，或在使用合并估计量时未检查两总体方差相等的假设。在进阶统计学 2 中，始终要陈述你所做的假设。
9. Estimator Bias and Variance | 估计量偏差与方差
An estimator θ̂ is unbiased if E(θ̂) = θ. When asked to find E(θ̂ dependence. A typical slip is claiming E(X²) = [E(X)]², which is true only for a constant.
若 E(θ̂) = θ，则估计量 θ̂ 是无偏的。求 E(θ̂) 时，许多考生会线性展开，却忘了 E(kX) = kE(X) 和 E(X+Y)=E(X)+E(Y) 无论是否独立都成立，而方差规则需要独立性。典型失误是声称 E(X²) = [E(X)]²，但这仅当 X 为常数时成立。
When comparing two unbiased estimators, the one with smaller variance is more efficient. Students sometimes confuse efficiency with bias or fail to compute Var(θ̂) correctly for linear combinations, particularly when the sample is drawn without replacement from a finite population.
比较两个无偏估计量时，方差更小者更有效。学生有时会把有效性与偏差搞混，或在为线性组合计算 Var(θ̂) 时出错，尤其当样本来自有限总体且为不放回抽样时。
10. General Formula Slips in FS2 | FS2 常见公式笔误
The formula sheet is a help, not a substitute for understanding. Watch out for: Sxy = Σ(x−x̄)(y−ȳ) = Σxy − (Σx Σy)/n – a missing division by n in the second term ruins results. Similarly, correlation r = Sxy / √Sxx Syy; some candidates take square root only of the denominator as a whole or mis‑apply brackets.
公式表是辅助，不能替代理解。要当心：Sxy = Σ(x−x̄)(y−ȳ) = Σxy − (Σx Σy)/n——第二项漏掉除以 n 会毁掉结果。类似地，相关系数 r = Sxy / √Sxx Syy；有些考生只对分母整体开方或括号使用不当。
ation: Z = (X−μ)/σ, not (X−μ)/σ². Similarly, for the sample mean, the standard error is σ/√n, not σ²/n. Always double‑check which measure of spread is required.
在连续分布中，概率等于面积。标准化时，一个极常见的失误是除以方差而非标准差：Z = (X−μ)/σ，而不是 (X−μ)/σ²。类似地，对于样本均值，标准误为 σ/√n，而非 σ²/n。务必反复检查需要的是哪种离散度量。
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