📚 Pre-U Edexcel Statistics: Cross-Disciplinary Integrated Question Training | Pre-U Edexcel 统计：跨学科综合题型训练
In Edexcel Pre-U Statistics, the ability to apply statistical methods across different disciplines is essential. This integrated training article presents typical cross-disciplinary problems from biology, economics, engineering, and more, showing how hypothesis testing, regression, probability distributions, and estimation are used in real-world contexts. Through paired worked examples, you will strengthen your conceptual understanding and exam technique.
在爱德思 Pre-U 统计学中，能够将统计方法应用于不同学科至关重要。本文综合训练涵盖了生物、经济、工程等领域的典型问题，展示假设检验、回归、概率分布及估计在真实情境中的应用。通过配对的例题讲解，你将加深概念理解并提升应试技巧。
1. Biology: Two-Sample t-Test and Confidence Interval | 生物学：双样本 t 检验与置信区间
A clinical trial compares systolic blood pressure reduction after taking two new drugs. Group 1 (n₁ = 12) has sample mean reduction x̄₁ = 10.5 mmHg with standard deviation s₁ = 2.8; Group 2 (n₂ = 10) has x̄₂ = 8.2 mmHg, s₂ = 3.1. Assume equal variances. Construct a 95% confidence interval for μ₁ – μ₂ and interpret.
一项临床试验比较两种新药对收缩压的降低效果。组1 (n₁=12) 的样本均值降低 x̄₁=10.5 mmHg，标准差 s₁=2.8；组2 (n₂=10) 的 x̄₂=8.2 mmHg，s₂=3.1。假设总体方差相等，构建 μ₁ – μ₂ 的 95% 置信区间并给出解释。
Pooled variance: sₚ² = [(n₁-1)s²₁ + (n₂-1)s²₂] / (n₁ + n₂ – 2) = [11×2.8² + 9×3.1²] / 20 = [86.24 + 86.49] / 20 = 8.6365. Standard error SE = sₚ √(1/n₁ + 1/n₂) = √8.6365 × √(1/12 + 1/10) ≈ 2.939 × 0.4282 ≈ 1.258. The t-critical value for df=20 at 95% is 2.086. CI = (10.5 – 8.2) ± 2.086×1.258 = 2.3 ± 2.624 → (-0.324, 4.924). Since the interval includes 0, there is no significant difference at the 5% level.
合并方差 sₚ² = [(n₁-1)s²₁ + (n₂-1)s²₂]/(n₁+n₂-2) = [11×2.8² + 9×3.1²]/20 = 8.6365。标准误 SE = sₚ√(1/n₁+1/n₂) ≈ 1.258。自由度为20的95% t临界值为2.086，置信区间为 (10.5-8.2) ± 2.086×1.258 = 2.3 ± 2.624，即 (-0.324, 4.924)。区间包含0，因此在5%水平下差异不显著。
2. Economics: Time Series Decomposition and Forecasting | 经济学：时间序列分解与预测
The quarterly sales of a company (in £1000) are recorded: Q1-2023=42, Q2=50, Q3=68, Q4=58. The moving average for the first four quarters is (42+50+68+58)/4 = 54.5. Calculate centred moving averages and estimate the seasonal component for Q3 using the additive model.
某公司季度销售额（千英镑）为：Q1-2023=42，Q2=50，Q3=68，Q4=58。前四个季度的移动平均为 (42+50+68+58)/4=54.5。计算中心化移动平均，并用加法模型估计Q3的季节成分。
After aligning centred moving averages with quarters, the raw seasonal effect for Q3 = actual – CMAtrend = 68 – 56.25 = 11.75. Using the simple average of all available Q3 deviations, we can produce a seasonal factor. Forecasting future values then involves the trend line and adding the appropriate seasonal factor. This method is widely used in economic planning and retail demand forecasting.
将中心化移动平均对齐到季度后，Q3的原始季节效应 = 实际值 – CMA趋势 = 68 – 56.25 = 11.75。利用所有可用Q3离差的简单平均，可得到季节因子。未来预测需要趋势线和加上相应季节因子。该方法广泛应用于经济规划和零售需求预测。
3. Engineering: Poisson Process and Reliability | 工程：泊松过程与可靠性
The number of defects in a manufactured microchip follows a Poisson distribution with mean λ = 0.5 per chip. Find the probability that a randomly chosen chip has zero defects, and the probability that in a batch of 10 chips at most one chip has at least one defect.
某微芯片的缺陷数服从泊松分布，均值为 λ=0.5 个/芯片。求随机选取一块芯片零缺陷的概率，以及在一批10块芯片中至多1块芯片出现至少一个缺陷的概率。
P(X=0) = e^(-0.5) × 0.5⁰ / 0! = e^(-0.5) ≈ 0.6065. For a single chip, P(at least one defect) = 1 – 0.6065 = 0.3935. The binomial count among 10 chips (success = chip has defect) is B(10, 0.3935). P(0 or 1 success) = C(10,0)(0.3935)⁰(0.6065)¹⁰ + C(10,1)(0.3935)¹(0.6065)⁹ ≈ 0.0061 + 0.0403 ≈ 0.0464. Engineering reliability heavily uses such Poisson–binomial cascades.
P(X=0)=e⁻⁰·⁵ ≈ 0.6065。单块芯片至少一个缺陷的概率为 1-0.6065=0.3935。10块中缺陷芯片数服从二项分布 B(10, 0.3935)。P(0或1成功) ≈ 0.0464。工程可靠性分析频繁使用这种泊松—二项连锁模型。
4. Psychology: Chi-Squared Test for Independence | 心理学：卡方独立性检验
A psychologist surveys 200 individuals on sleep quality (poor, average, good) and exercise habit (regular, occasional, none). The observed frequencies are given. Perform a χ² test at the 5% significance level to determine if sleep quality and exercise are associated.
一位心理学家调查了200人的睡眠质量（差、一般、好）与运动习惯（定期、偶尔、从不）。给定观测频数，在5%显著性水平下进行χ²检验，判断睡眠质量与运动是否有关联。
Expected frequencies are calculated under independence: eᵢⱼ = (row total × column total) / grand total. The test statistic χ² = Σ (oᵢⱼ – eᵢⱼ)² / eᵢⱼ is compared against the critical value with (r-1)(c-1) degrees of freedom. In this case df = (3-1)(3-1)=4, χ²_crit ≈ 9.488. If χ² exceeds the critical value, we reject the null hypothesis of independence, suggesting a link between lifestyle and sleep. This is a standard tool in psychological surveys.
期望频数在独立假设下计算：eᵢⱼ=(行合计×列合计)/总计。检验统计量 χ²=Σ (oᵢⱼ – eᵢⱼ)²/eᵢⱼ 与自由度为 (r-1)(c-1)=4 的临界值约9.488比较。若χ²大于临界值，拒绝独立零假设，提示生活方式与睡眠有关联。这是心理学问卷中的标准工具。
5. Environmental Science: Stratified Sampling and Estimation | 环境科学：分层抽样与估计
A forestry survey divides a forest into three strata: young, mature, old trees. Using proportional allocation, the mean height (m) and standard deviation are measured in each stratum. Construct an unbiased estimate of the overall mean height and its standard error.
某林业调查将一片森林分为幼树、中龄树、老树三个层。按比例分配抽样，每层测得了平均高度（米）和标准差。构建总体平均高度的无偏估计及其标准误。
If stratum sizes are N₁=400, N₂=250, N₃=100 with sample sizes n₁=40, n₂=25, n₃=10, and sample means x̄₁=8.2, x̄₂=14.5, x̄₃=22.1, variances s²₁= 2.1, s²₂=3.4, s²₃=4.8, the overall mean estimate is x̄ₛₜ = Σ (Nᵢ/N) x̄ᵢ = (400/750)×8.2 + (250/750)×14.5 + (100/750)×22.1 ≈ 12.24 m. The standard error SE = √[ Σ (Nᵢ/N)² × (1 – nᵢ/Nᵢ) × s²ᵢ / nᵢ ]. Such techniques underpin environmental resource inventories.
各层大小 N₁=400, N₂=250, N₃=100，样本量 n₁=40, n₂=25, n₃=10，样本均值 x̄₁=8.2, x̄₂=14.5, x̄₃=22.1，方差 s²₁=2.1, s²₂=3.4, s²₃=4.8，总体均值估计 x̄ₛₜ = Σ (Nᵢ/N)x̄ᵢ ≈ 12.24 m。标准误 SE = √[ Σ (Nᵢ/N)² (1-nᵢ/Nᵢ) s²ᵢ/nᵢ ]。这类技术是环境资源清查的基础。
6. Medicine: Normal Distribution and Reference Ranges | 医学：正态分布与参考范围
Fasting blood glucose levels in healthy adults are approximately normally distributed with μ = 90 mg/dL and σ = 12 mg/dL. A diagnostic cut-off for pre-diabetes is set at the 97.5th percentile. Determine this threshold, and find the probability that a patient with true mean 105 mg/dL is classified as ‘normal’.
健康成人空腹血糖近似服从正态分布，μ=90 mg/dL，σ=12 mg/dL。糖尿病前期的诊断截点设在97.5分位数。确定该阈值，并计算真实均值为105 mg/dL的患者被归为“正常”的概率。
The 97.5th percentile = μ + 1.96σ = 90 + 1.96×12 = 113.52 mg/dL. For a patient whose true blood glucose is N(105, 12²), the probability of falling below 113.52 is P(Z < (113.52-105)/12) = P(Z < 0.71) ≈ 0.7611. Thus about 76% of such pre-diabetic patients would be misclassified as normal, illustrating the sensitivity-specificity trade-off in medical screening.
97.5分位数 = μ + 1.96σ = 113.52 mg/dL。对于真实血糖服从 N(105,12²) 的患者，其值低于113.52的概率为 P(Z < 0.71) ≈ 0.7611。因此约76%的这类糖尿病前期患者会被错分为正常，体现了医学筛查中灵敏度与特异度的权衡。
7. Physics & Quality Control: Normal Approximation to the Binomial | 物理与质量控制：二项分布的正态近似
A factory produces electrical resistors, and 5% are defective. A sample of 500 resistors is inspected. Use a normal approximation to find the probability that 30 or more are defective, applying a continuity correction.
一家工厂生产电阻，5%为次品。检验了500个电阻的样本。运用正态近似并加连续性校正，求次品数不少于30的概率。
The number of defectives X ~ Bin(500, 0.05). Mean μ = np = 25, variance σ² = np(1-p) = 23.75, σ ≈ 4.873. Applying a continuity correction, P(X ≥ 30) ≈ P(Z ≥ (29.5 – 25) / 4.873) = P(Z ≥ 0.923) ≈ 0.178. Without correction the approximation loses accuracy; such corrections are standard in industrial quality monitoring and physics experiments where large binomial counts appear.
次品数 X ~ Bin(500, 0.05)。均值 μ=np=25，方差 σ²=np(1-p)=23.75，σ≈4.873。连续性校正后，P(X≥30) ≈ P(Z ≥ (29.5-25)/4.873) = P(Z ≥ 0.923) ≈ 0.178。未校正会损失精度；这类校正在工业质量监控和出现大计数二项分布变量的物理实验中是标准操作。
8. Social Sciences: Linear Regression Interpretation | 社会科学：线性回归解释
A study on education and income produces the regression equation: Predicted income = 12 000 + 2 500 × years_of_education, with R² = 0.42. Interpret the slope, the intercept, and the meaning of R² in the context of social research. Also discuss why correlation does not imply causation.
一项关于教育年限与收入的研究给出回归方程：预测收入 = 12 000 + 2 500 × 受教育年限，R²=0.42。从社会研究的角度解释斜率、截距和 R² 的含义，并讨论为什么相关不等于因果。
The slope indicates that each additional year of education is associated with an average increase of £2 500 in annual income. The intercept of £12 000 estimates the baseline income for someone with zero years of formal education, though extrapolation may be unreliable. R²=0.42 means 42% of the variation in income is explained by years of education alone. However, confounding variables such as family background or ability could drive both education and income, meaning correlation is not sufficient to establish causality. In social science questions, always assess potential confounders.
斜率表示受教育年限每增加一年，平均年收入增加2500英镑。截距12000英镑估计了零正规教育年限者的基准收入，但外推可能不可靠。R²=0.42意味着仅受教育年限就解释了收入42%的变异。但混杂变量如家庭背景或能力可能同时影响教育与收入，因此相关不足以建立因果。社会科学题目中，始终要评估潜在混杂因素。
9. Finance: Log-Normal Distribution and Value at Risk | 金融：对数正态分布与风险价值
An investment portfolio has daily returns that follow a log-normal distribution: ln(1+R) ~ N(0.0005, 0.02²). Calculate the 5% 1-day Value at Risk (VaR) for a £1 million portfolio, i.e. the maximum loss with 95% confidence.
某投资组合的日收益率服从对数正态分布：ln(1+R) ~ N(0.0005, 0.02²)。计算一个100万英镑组合的5% 1天风险价值(VaR)，即95%置信度下的最大损失。
The 5th percentile of the normal distribution is μ – 1.645σ for the log-return: z = 0.0005 – 1.645×0.02 = -0.0324. Therefore, the corresponding simple return is R* = e^(z) – 1 = e⁻⁰·⁰³²⁴ – 1 ≈ -0.0319. VaR = £1 000 000 × 0.0319 = £31 900. This means there is a 5% chance that the portfolio will lose more than £31 900 in one day. Such calculations are integral to financial risk management and use continuous distributions learned in Pre-U statistics.
正态分布的第5百分位数对于对数收益率 z = 0.0005 – 1.645×0.02 = -0.0324。则简单收益率 R* = e^z – 1 ≈ -0.0319。VaR = 1 000 000×0.0319 = 31 900 英镑，意味着有5%的可能性组合一天内损失超过31 900英镑。这类计算是金融风险管理的核心，并运用了Pre-U统计学中的连续分布。
10. Genetics: Bayes’ Theorem and Probability Trees | 遗传学：贝叶斯定理与概率树
A certain recessive genetic condition occurs in 1 out of 10 000 births. A screening test has 99% sensitivity (detects condition if present) and 99.5% specificity (negative if healthy). If a newborn tests positive, what is the probability they actually have the condition? Use Bayes’ theorem.
某隐性遗传病在新生儿中发生率为 1/10 000。一项筛查试验灵敏度为99%（有病时检出阳性），特异度为99.5%（健康时检出阴性）。若某新生儿检测呈阳性，他实际患病的概率是多少？使用贝叶斯定理。
Let D be the event disease present. P(D)=0.0001, P(Pos|D)=0.99, P(Pos|D’)=1-0.995=0.005. Bayes’ theorem: P(D|Pos) = [P(Pos|D)P(D)] / [P(Pos|D)P(D) + P(Pos|D’)P(D’)] = (0.99×0.0001) / (0.99×0.0001 + 0.005×0.9999) = 0.000099 / (0.000099 + 0.0049995) = 0.000099 / 0
Published by TutorHao | Pre-U 统计 Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导