Pre-U Cambridge Statistics: Interdisciplinary Comprehensive Problem-Solving Training

📚 Pre-U Cambridge Statistics: Interdisciplinary Comprehensive Problem-Solving Training | 跨学科综合题型训练

Interdisciplinary questions in the Cambridge Pre-U Statistics examination blend core statistical concepts with real-world contexts drawn from biology, economics, physics, and the social sciences. These problems demand not only computational accuracy but also the ability to select an appropriate model, justify assumptions, and interpret results in plain language. This article provides structured training across ten typical cross-curricular scenarios, reinforcing techniques such as hypothesis testing, regression, goodness-of-fit, survival analysis, and ANOVA while explicitly linking theory to authentic applications.

剑桥 Pre-U 统计考试中的跨学科题目将核心统计概念与生物学、经济学、物理学和社会科学等真实情境深度融合。这些题目不仅要求计算精确，更考验考生选择适当模型、论证假设并用通俗语言解释结果的能力。本文通过十个典型的跨学科场景进行结构化训练，强化假设检验、回归、拟合优度、生存分析和方差分析等技巧，同时将理论明确联系到实际应用。

1. Biology: Drug Efficacy and Two-Sample Tests | 生物学：药物有效性与双样本检验

A pharmaceutical trial measures systolic blood pressure reduction in 28 patients randomly assigned to a new drug (Group A) and a placebo (Group B). The sample means are x̄ₐ = 18.2 mmHg and x̄_B = 9.4 mmHg, with unbiased variance estimates sₐ² = 38.5 and s_B² = 42.1. Before conducting a formal test, you must verify that the populations are approximately normal and check the assumption of equal variances using an F-test. The practical question is whether the new drug produces a significantly greater mean reduction at the 5% level.

某药物试验测量了 28 位患者的收缩压降幅，患者被随机分配至新药组（A 组）与安慰剂组（B 组）。样本均值分别为 x̄ₐ = 18.2 mmHg 与 x̄_B = 9.4 mmHg，无偏方差估计值为 sₐ² = 38.5 与 s_B² = 42.1。在正式检验前，你必须验证总体近似正态，并通过 F 检验考察方差齐性假设。实际要回答的问题是：在 5% 显著性水平下，新药是否带来了显著更大的平均降幅。

Assuming equal variances, the pooled variance is sₚ² = ( (nₐ–1)sₐ² + (n_B–1)s_B² )/(nₐ+n_B–2). With nₐ = n_B = 14, the test statistic is t = (x̄ₐ – x̄_B) / √(sₚ²(1/nₐ + 1/n_B)). The result is compared against a t-distribution with 26 degrees of freedom. If the p-value falls below 0.05, we reject H₀: μₐ = μ_B and conclude a significant difference. This classic independent-samples design illustrates how Pre-U candidates must integrate F-tests, t-tests, and contextual interpretation within a biomedical setting.

假设方差相等，合并方差为 sₚ² = ( (nₐ–1)sₐ² + (n_B–1)s_B² )/(nₐ+n_B–2)。由于 nₐ = n_B = 14，检验统计量为 t = (x̄ₐ – x̄_B) / √(sₚ²(1/nₐ + 1/n_B))。结果与自由度为 26 的 t 分布比较。若 p 值小于 0.05，则拒绝原假设 H₀: μₐ = μ_B，得出显著差异的结论。这一经典独立样本设计展现了 Pre-U 考生如何在生物医学背景下综合运用 F 检验、t 检验和情境化解读。

2. Economics: Consumption-Income Regression | 经济学：消费-收入回归分析

An economist collects annual household data on disposable income x (thousand £) and consumption expenditure y (thousand £) for 40 randomly selected families. The least squares line is fitted as y = 2.15 + 0.78x, with a residual standard error of 1.42. The sample correlation coefficient is r = 0.891. The analyst needs to test whether the slope is significantly different from zero and construct a 95% confidence interval for the marginal propensity to consume.

某经济学家收集了 40 个随机家庭的年可支配收入 x（千英镑）与消费支出 y（千英镑）数据。最小二乘直线拟合为 y = 2.15 + 0.78x，残差标准误为 1.42。样本相关系数 r = 0.891。分析者需要检验斜率是否显著异于零，并构建边际消费倾向的 95% 置信区间。

The t-test for the slope uses t = (b̂ – 0) / SE(b̂), where SE(b̂) = √[ (residual SS/(n–2)) / Σ(xᵢ – x̄)² ]. With n = 40, degrees of freedom = 38. The strong positive r also allows an alternative test using the t-transformation of r. The confidence interval is b̂ ± t_crit × SE(b̂). Interpreting the interval in economic terms—for instance, “a £1000 increase in disposable income is associated with an increase in consumption of between £726 and £834″—bridges statistical output and economic reasoning, a skill frequently assessed in Pre-U papers.

斜率的 t 检验使用 t = (b̂ – 0) / SE(b̂)，其中 SE(b̂) = √[ (残差平方和/(n–2)) / Σ(xᵢ – x̄)² ]。n = 40，自由度为 38。较强的正相关 r 也可通过 r 的 t 变换进行替代检验。置信区间为 b̂ ± t_crit × SE(b̂)。以经济学语言解读区间——“可支配收入每增加 1000 英镑，消费支出增加约 726 至 834 英镑”——在统计输出和经济论证之间架起桥梁，这是 Pre-U 试卷中经常评估的能力。

3. Physics: Poisson Goodness-of-Fit for Radioactive Decay | 物理学：放射性衰变的泊松拟合优度

A student measures the number of alpha particles emitted from a weak polonium source in 250 fixed 10-second intervals. The observed frequencies of counts 0, 1, 2, 3, 4+ are 57, 86, 61, 32, and 14. The mean count per interval is 1.52. The student wishes to test at the 1% significance level whether the emissions follow a Poisson distribution, which would validate the assumption of independent random decay events.

一名学生测量了弱钋源在 250 个固定 10 秒间隔内发射的 α 粒子数。计数 0, 1, 2, 3, 4+ 的观测频数分别为 57, 86, 61, 32, 14。每间隔平均计数为 1.52。学生希望在 1% 显著性水平下检验发射数是否服从泊松分布，以验证随机独立衰变假设。

The expected frequencies are computed using Poisson probabilities P(X = k) = (e^–λ λ^k)/k! with λ = 1.52. Combining the tail (4+) ensures all expected frequencies exceed 5. The test statistic is χ² = Σ (O – E)² / E. Degrees of freedom are (number of cells – 1 – number of estimated parameters), here 5 – 1 – 1 = 3. If χ² > χ²_crit(3, 0.01), the Poisson model is rejected, implying the need for a more complex model like a mixture or compound distribution. This modern physics context demands careful handling of parameter estimation and the χ² approximation.

期望频数通过泊松概率 P(X = k) = (e^–λ λ^k)/k! 计算，其中 λ = 1.52。将尾部（4+）合并以确保所有期望值大于 5。检验统计量为 χ² = Σ (O – E)² / E。自由度为（组数 – 1 – 估计参数个数），此处为 5 – 1 – 1 = 3。若 χ² > χ²_crit(3, 0.01)，则拒绝泊松模型，表明可能需要混合或复合分布等更复杂模型。这一近代物理情境要求考生谨慎处理参数估计和卡方近似。

4. Environmental Science: Stratified Sampling of Water Quality | 环境科学：水质分层的抽样设计

An environmental agency monitors nitrate levels (mg/L) in a river network with three distinctive zones: upstream, industrial, and downstream. The strata areas are 80 km², 60 km², and 110 km². The agency decides on proportional allocation with a total sample size of 75 water samples. Prior pilot data give stratum standard deviations σ₁ = 2.1, σ₂ = 4.5, σ₃ = 3.0. The goal is to estimate the overall mean concentration and its standard error, and then compute a 95% confidence interval.

某环境机构监测一条河流网络上、中（工业区）、下游三个区域的水中硝酸盐浓度 (mg/L)。各层面积分别为 80 km²、60 km² 和 110 km²。机构决定采用比例分配，共抽取 75 个水样。前期试点数据给出各层标准差 σ₁ = 2.1、σ₂ = 4.5、σ₃ = 3.0。目标是估计总体平均浓度及其标准误，并计算 95% 置信区间。

The stratum sample sizes under proportional allocation are n_h = n × (N_h/N), where N_h is the area weight. The estimate of the population mean is x̄_st = Σ w_h x̄_h. The standard error is SE(x̄_st) = √[ Σ w_h² × (s_h²/n_h) ], using finite population correction if necessary. The resulting interval shows the precision gained by stratification: the heavy-pollution midstream gets more representation, reducing overall variance. Linking sampling theory to geographic stratification develops a crucial Pre-U skill—choosing appropriate designs for real ecological studies.

比例分配下的各层样本量为 n_h = n × (N_h/N)，N_h 为面积权重。总体均值估计为 x̄_st = Σ w_h x̄_h。标准误为 SE(x̄_st) = √[ Σ w_h² × (s_h²/n_h) ]，必要时使用有限总体校正。所得区间显示出分层带来的精度提升：污染严重的中游获得更多代表性，降低了总方差。将抽样理论与地理分层联系起来，培养了一项关键 Pre-U 技能——为真实生态研究选择合适设计。

5. Psychology: Paired t-Test for Memory Scores | 心理学：记忆分数的配对 t 检验

Twenty participants are tested under two conditions: immediate recall (A) and delayed recall after 24 hours (B). The differences d = A – B have a mean of d̄ = 7.3 words and a standard deviation of s_d = 4.8. The psychologist wants to know if immediate recall scores are significantly higher, assuming normality of differences. Both familiar and unfamiliar word lists are used, adding a blocking factor to control for vocabulary familiarity.

20 名被试在两种条件下接受测试：即时回忆 (A) 和 24 小时后延迟回忆 (B)。差值 d = A – B 的均值 d̄ = 7.3 个单词，标准差 s_d = 4.8。心理学家想知道即时回忆分数是否显著更高，假设差值服从正态分布。实验中同时使用了熟悉词与不熟悉词列表，增加了一个区组因素以控制词汇熟悉度。

The paired t-test statistic is t = d̄ / (s_d/√n), with n = 20. Against the one-sided alternative H₁: μ_d > 0, the critical value at α = 0.05 is t_19,0.05. If the assumption of normality is doubted, the Wilcoxon signed-rank test can be applied as a non-parametric alternative. This design exemplifies how psychological experiments reduce variability through blocking and how Pre-U candidates must justify the choice between parametric and non-parametric procedures based on data diagnostics.

配对 t 检验统计量为 t = d̄ / (s_d/√n)，n = 20。对于单侧备择假设 H₁: μ_d > 0，α = 0.05 时的临界值为 t_19,0.05。若对正态性假设存疑，可采用 Wilcoxon 符号秩检验作为非参数替代。该设计展示了心理学实验如何通过区组化降低变异性，以及 Pre-U 考生如何基于数据诊断合理选择参数或非参数方法。

6. Medicine: Survival Analysis Using Exponential Models | 医学：利用指数模型进行生存分析

A clinical trial follows 50 patients receiving a new cancer therapy. The total observed follow-up time is 420 person-months, and 22 patients experienced the event of interest. If we assume the survival time follows an exponential distribution with constant hazard rate λ, the maximum likelihood estimate is λ̂ = number of events / total person-time = 22/420 ≈ 0.0524 per month. The six-month survival probability is estimated as Ŝ(6) = e^–λ̂×6 ≈ 0.73.

一项临床试验跟踪了 50 名接受新型癌症疗法的患者。总观察随访时间为 420 人月，22 名患者经历了目标事件。若假定生存时间服从恒定风险率 λ 的指数分布，最大似然估计为 λ̂ = 事件数 / 总人时 = 22/420 ≈ 0.0524 每月。六个月生存概率估计为 Ŝ(6) = e^–λ̂×6 ≈ 0.73。

To construct a 95% confidence interval for S(6), one first finds an interval for λ using the asymptotic normality of λ̂: SE(λ̂) = √(λ̂ / total person-time) ≈ 0.0112, giving λ ± 1.96 × SE. Then transform the endpoints via e^–6λ. The comparison between two therapies can be formalized with a log-rank test or by fitting an exponential model with a treatment indicator. Handling censored data and interpreting hazard rates are central to Pre-U medical statistics questions.

为构建 S(6) 的 95% 置信区间，首先利用 λ̂ 的渐近正态性求得 λ 的区间：SE(λ̂) = √(λ̂ / 总人时) ≈ 0.0112，得到 λ ± 1.96 × SE。然后通过 e^–6λ 变换端点。两种疗法的比较可通过 Log-rank 检验或拟合含治疗指示变量的指数模型规范化。处理删失数据并解读风险率是 Pre-U 医学统计题的核心。

7. Business: Confidence Interval for a Proportion in Market Research | 商业：市场调查中比例的置信区间

A company surveys 800 potential customers and finds that 360 are aware of its new brand. The business needs a 99% confidence interval for the true brand awareness proportion p. The sample proportion is p̂ = 360/800 = 0.45. The conditions for a large-sample interval are met because np̂ = 360 and n(1–p̂) = 440 are both well above 10.

某公司调查 800 名潜在客户，发现 360 名知晓其新品牌。企业需要真实品牌知名度比例 p 的 99% 置信区间。样本比例 p̂ = 360/800 = 0.45。大样本区间条件成立，因为 np̂ = 360 与 n(1–p̂) = 440 均远大于 10。

The standard error is SE = √[ p̂(1–p̂) / n ] = √(0.45×0.55/800) ≈ 0.0176. The z-multiplier for 99% is 2.576. So the interval is p̂ ± 2.576 × 0.0176, roughly (0.405, 0.495). Business decisions such as campaign continuation or product redesign hinge on whether this interval exceeds management’s target threshold, illustrating how statistical inference feeds directly into commercial strategy.

标准误为 SE = √[ p̂(1–p̂) / n ] = √(0.45×0.55/800) ≈ 0.0176。99% 置信水平的 z 乘数为 2.576，因此区间为 p̂ ± 2.576 × 0.0176，约 (0.405, 0.495)。诸如是否继续投放广告或重新设计产品等商业决策都取决于该区间是否超过管理层的目标阈值，体现了统计推断直接服务于商业策略的思维。

8. Sociology: Chi-Squared Test of Independence in a Contingency Table | 社会学：列联表的卡方独立性检验

A sociologist collects data on 600 adults, classifying them by education level (high school, bachelor, postgraduate) and voting intention (Party X, Party Y, Undecided). The observed table is:

Highest Edu	Party X	Party Y	Undecided	Total
High School	68	94	38	200
Bachelor	112	86	52	250
Postgraduate	70	45	35	150
Total	250	225	125	600

The research question is whether education level and voting intention are independent.

某社会学家收集了 600 名成年人的数据，按教育水平（高中、本科、研究生）和投票意向（X 党、Y 党、未决定）分类。观察频数如表所示。研究问题是教育水平与投票意向是否独立。

Expected frequencies are calculated as (row total × column total)/grand total. The χ² statistic = Σ (O – E)²/E, with degrees of freedom (3–1)×(3–1) = 4. If the computed χ² exceeds the critical value χ²₄(0.05), we reject the null hypothesis of independence. To measure the strength of association, Cramér’s V is often required. This sociological example trains candidates to communicate the societal implications of a significant association, a nuanced skill far beyond mere computation.

期望频数按（行合计 × 列合计）/ 总计计算。χ² 统计量 = Σ (O – E)²/E，自由度为 (3–1)×(3–1) = 4。若计算所得 χ² 超过临界值 χ²₄(0.05)，则拒绝独立性原假设。通常还会要求计算 Cramér’s V 以衡量关联强度。这一社会学例子训练考生传达显著关联的社会含义，这远非纯粹计算，而是更精细的技能。

9. Finance: Testing Normality of Asset Returns | 金融：资产收益的正态性检验

A quantitative analyst reviews daily log-returns of a stock index over 250 trading days. The sample skewness is –0.52 and the sample excess kurtosis is 1.14. Because many financial models (e.g., the Black-Scholes framework) assume normally distributed returns, the analyst performs a Jarque-Bera test. The test statistic is JB = n [(skewness²)/6 + (excess kurtosis²)/24].

一位量化分析师审视了某股指 250 个交易日的日对数收益率。样本偏度为 –0.52，样本超额峰度为 1.14。由于许多金融模型（如 Black-Scholes 框架）假设收益率服从正态分布，分析师执行 Jarque-Bera 检验。检验统计量为 JB = n [(偏度²)/6 + (超额峰度²)/24]。

With n = 250, JB = 250 × [ (–0.52)²/6 + (1.14)²/24 ] = 250 × (0.045 + 0.054) = 24.8. Under the null of normality, JB follows a χ² distribution with 2 degrees of freedom. The p-value is extremely small, so normality is rejected. This makes the analyst consider heavier-tailed distributions such as the t-distribution with low df. This finance scenario integrates moment-based statistics, χ² tables, and model selection—a typical Pre-U synoptic question.

由 n = 250，JB = 250 × [ (–0.52)²/6 + (1.14)²/24 ] = 250 × (0.045 + 0.054) = 24.8。在正态性原假设下，JB 服从自由度为 2 的 χ² 分布。p 值极小，故拒绝正态性。这促使分析师考虑厚尾分布，如低自由度的 t 分布。该金融场景整合了基于矩的统计量、χ² 查表和模型选择，是典型的 Pre-U 综述性题目。

10. Public Health: Two-Way ANOVA for Intervention Studies | 公共健康：干预研究的双因素方差分析

A public health trial assesses the impact of two exercise programmes (A and B) and two diet plans (Low-carb and Mediterranean) on BMI reduction after 12 weeks. Eight participants are assigned to each of the four combinations (32 in total). The researchers record BMI reduction and aim to evaluate main effects and possible interaction. Summary statistics per cell show varying means and a hint of non-additivity.

一项公共健康试验评估两种运动方案（A 和 B）与两种饮食计划（低碳水与地中海饮食）在 12 周后对 BMI 降幅的影响。每种组合分配 8 名参与者（共 32 人）。研究人员记录 BMI 降幅，旨在评估主效应和可能的交互作用。各细格的汇总统计量显示均值不等，并暗示非可加性。

The two-way ANOVA model is x_ijk = μ + α_i + β_j + (αβ)_ij + ε_ijk. Sums of squares for Exercise, Diet, Interaction, and Error are computed, each with its own degrees of freedom. The F-ratio for interaction tests whether the effect of diet depends on the exercise programme. If significant, simple effects are examined. Reporting the output in a clear table and interpreting the public health meaning—such as recommending the combination of Programme B with Mediterranean diet—concludes this rigorous statistical consulting simulation.

双因素方差分析模型为 x_ijk = μ + α_i + β_j + (αβ)_ij + ε_ijk。计算运动、饮食、交互作用及误差的平方和，各有其自由度。交互作用的 F 比检验饮食效果是否依赖于运动方案。若交互作用显著，则进一步考察简单效应。以清晰表格呈现输出，并解读公共健康含义——例如推荐 B 方案结合地中海饮食——完成了这一严谨的统计咨询模拟。

Published by TutorHao | Statistics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Pre-U Cambridge Statistics: Interdisciplinary Comprehensive Problem-Solving Training | 跨学科综合题型训练