Interdisciplinary Integrated Problem Training for Pre-U WJEC Statistics | Pre-U WJEC 统计：跨学科综合题型训练

📚 Interdisciplinary Integrated Problem Training for Pre-U WJEC Statistics | Pre-U WJEC 统计：跨学科综合题型训练

The Pre-U WJEC Statistics syllabus demands more than isolated technique drills; it requires learners to select, justify, and synthesise statistical methods within genuine interdisciplinary scenarios. This article presents a structured collection of integrated problems drawn from biology, economics, psychology, engineering, medicine, environmental science, business, and sociology. Each section unpacks a complete contextual question, highlighting how core statistical tools – from Poisson models and t‑tests to survival analysis and Bayesian updating – are deployed when the data carry subject‑specific meaning. Working through these exercises will deepen conceptual understanding and sharpen the transferable analytical skills essential for high achievement in the final assessment.

Pre-U WJEC 统计课程要求考生不仅掌握单一技巧，更要能在真实的跨学科情境中选择、论证并综合运用统计方法。本文提供一套结构化的综合题型训练，取材自生物学、经济学、心理学、工程学、医学、环境科学、商业和社会学等领域。每个小节围绕一道完整的背景题展开，重点展示泊松模型、t 检验、生存分析、贝叶斯更新等核心工具如何在学科意义上被激活。通过逐题精练，学生将深化概念理解，磨炼可迁移的分析能力，这正是拿下高分的必备素养。

1. Biology: Poisson Distribution for Bacterial Counts | 生物学：细菌计数的泊松分布

A food safety laboratory tests 50 independent water samples, each of volume 1 mL, and records the number of bacterial colonies per sample. The historical mean count is known to be 2.0 colonies per mL under clean conditions. The technician suspects recent contamination has raised the true mean. The data yield a sample total of 127 colonies across the 50 samples. State appropriate hypotheses, explain why a Poisson model is suitable, and carry out a hypothesis test at the 5% significance level. Discuss whether a normal approximation is justified.

某食品安全实验室检测了 50 份独立的 1 mL 水样，记录每份样品的细菌菌落数。已知清洁条件下的历史均值是每毫升 2.0 个菌落。技术员怀疑近期污染导致真实均值上升。50 份样品的菌落总数为 127。请写出相应的假设，说明为何泊松模型适用，并在 5% 显著性水平下进行假设检验。讨论正态近似是否合理。

A Poisson distribution is the natural model for counts of rare, independent events occurring in a fixed continuous unit, such as bacterial colonies in a fixed volume of liquid. Each 1 mL sample can be seen as an independent realisation of a Poisson process with rate λ. The hypotheses are H₀: λ = 2 against H₁: λ > 2. The total count T = 127 across n = 50 samples serves as the test statistic. Under H₀, T ~ Poisson(nλ₀) = Poisson(100). Because the mean 100 is large, the normal approximation to the Poisson is justified: T ~ N(100, 100) approximately. The observed test statistic gives z = (127 − 100)/√100 = 2.70. The one‑tailed critical value at 5% is 1.645. Since 2.70 > 1.645, we reject H₀ and conclude there is sufficient evidence of increased contamination.

泊松分布是描述固定连续单位内稀有独立事件出现次数的自然模型，恰好吻合固定体积液体中细菌菌落数的情境。每份 1 mL 样品可视为速率为 λ 的泊松过程的一次独立实现。假设为 H₀: λ = 2，H₁: λ > 2。检验统计量为 50 份样品的总菌落数 T = 127。在 H₀ 下，T ~ Poisson(100)。由于均值 100 较大，可用正态近似：T 近似服从 N(100, 100)。观测 z = (127 − 100)/√100 = 2.70，单尾 5% 临界值 1.645。因 2.70 > 1.645，拒绝 H₀，有充分证据表明污染均值上升。解此题时必须注意独立抽样假设和泊松聚合性质，并清楚交代正态近似成立的条件。

2. Economics: Correlation and Regression for Market Data | 经济学：市场数据的相关与回归

An economist collects annual data on consumer spending (in £ billions, denoted y) and disposable income (x) for a country over 12 consecutive years. The summary statistics are: Σx = 158.4, Σy = 131.6, Σx² = 2150.7, Σy² = 1493.4, Σxy = 1789.1. (a) Calculate Pearson’s product‑moment correlation coefficient r and interpret its value. (b) Find the equation of the least‑squares regression line of y on x. (c) Predict the spending when income is 14.0 and comment on the reliability of the prediction. (d) Test whether the true slope β₁ is significantly different from zero at the 1% level.

某经济学家收集了一国连续 12 年的年度消费者支出 y（单位：十亿英镑）与可支配收入 x 的数据。汇总统计量为：Σx = 158.4，Σy = 131.6，Σx² = 2150.7，Σy² = 1493.4，Σxy = 1789.1。(a) 计算皮尔逊积矩相关系数 r 并解释其数值；(b) 求 y 对 x 的最小二乘回归直线方程；(c) 预测收入为 14.0 时的支出，并评论预测的可靠性；(d) 在 1% 水平下检验总体斜率 β₁ 是否显著异于零。

Using the formula r = (nΣxy − ΣxΣy) / √{[nΣx² − (Σx)²][nΣy² − (Σy)²]}, substitution gives r ≈ 0.972. This extremely high positive correlation suggests a strong linear association between income and spending. The regression line is y = a + bx, with b = (nΣxy − ΣxΣy)/(nΣx² − (Σx)²) ≈ 0.884 and a = ȳ − b x̄ ≈ 1.54. Hence ŷ = 1.54 + 0.884x. For x = 14.0, ŷ = 13.92. The prediction lies within the observed range of income (interpolation is safer), but one must check residuals for homoscedasticity and normality. To test H₀: β₁ = 0 against H₁: β₁ ≠ 0, compute the standard error of b from the residual sum of squares. With a t‑statistic far exceeding the 1% critical value for 10 degrees of freedom, we reject H₀, confirming income is a significant predictor. These tasks mirror exactly the WJEC Pre‑U emphasis on using summary data and interpreting output in economic policy contexts.

代入公式 r = (12×1789.1 − 158.4×131.6) / √{[12×2150.7 − 158.4²][12×1493.4 − 131.6²]}，得 r ≈ 0.972，极强正相关表明收入与支出之间存在紧密的线性关系。回归直线 y = 1.54 + 0.884x。当 x = 14.0 时，预测支出为 13.92（十亿）。预测值落在观测区域内，插值可靠度较高，但应检查残差是否满足同方差性和正态性。检验 H₀: β₁ = 0 时，b 的标准误由剩余平方和导得，t 统计量远超自由度为 10 的 1% 临界值，拒绝 H₀，说明收入对支出有显著预测力。这类练习完全对应 WJEC Pre‑U 要求，即利用汇总统计量并解释经济政策含义。

3. Psychology: Independent Samples t‑Test | 心理学：独立样本的 t 检验

A researcher compares the effectiveness of two therapeutic techniques on reducing anxiety. Group A (n₁ = 15) receives cognitive‑behavioural therapy, while Group B (n₂ = 15) receives mindfulness training. After 8 weeks, anxiety scores are recorded. The sample means are x̄₁ = 24.6, x̄₂ = 30.1; the sample variances are s²₁ = 18.3, s²₂ = 22.9. Assume equal population variances. (a) Perform an independent‑samples t‑test to determine if there is a significant difference in mean anxiety scores at the 5% significance level. (b) Construct a 95% confidence interval for the difference in population means. (c) Discuss the assumptions and suggest what to do if the equal‑variance assumption is violated.

一位研究者比较两种治疗方法降低焦虑的效果。A 组（15 人）接受认知行为疗法，B 组（15 人）接受正念训练。8 周后记录焦虑分数：x̄₁ = 24.6，x̄₂ = 30.1；样本方差 s²₁ = 18.3，s²₂ = 22.9。假定总体方差相等。(a) 进行独立样本 t 检验，在 5% 显著性水平下判断平均焦虑分数是否存在显著差异；(b) 构建总体均值差的 95% 置信区间；(c) 讨论相关假定，并说明方差不等时应如何处理。

Under the equal‑variance assumption, the pooled variance is s²_p = [(14×18.3 + 14×22.9) / 28] = 20.6. The standard error of the difference in means is √(s²_p/15 + s²_p/15) = √(2×20.6/15) ≈ 1.656. The test statistic is t = (24.6 − 30.1)/1.656 = −3.32. For a two‑tailed test at 5% with 28 degrees of freedom, the critical value is approximately 2.048. Since |−3.32| > 2.048, we reject H₀ and conclude there is a significant difference. The 95% confidence interval is (−5.5 ± 2.048×1.656) = (−8.89, −2.11). This interval does not contain zero, consistent with the test outcome. Key assumptions include normality of scores in each population and independence within and between groups. If variances are unequal, Welch’s t‑test should be used, which adjusts the degrees of freedom downward. In Pre‑U examinations, candidates must be ready to select and justify the appropriate version of the t‑test.

在方差齐性假定下，合并方差 s²_p = 20.6，均值差的标准误为 1.656。t = −3.32。对自由度为 28 的双尾检验，5% 临界值约 2.048。|t| 大于临界值，拒绝 H₀，存在显著差异。95% 置信区间为 (−8.89, −2.11)，不含零，与检验结论一致。核心假定包括每组总体正态性和组内、组间的独立性。若方差不齐，则应采用 Welch’s t 检验并修正自由度。在 Pre‑U 考试中，考生必须能够根据背景选择并论证恰当的 t 检验版本。

4. Engineering: Exponential Distribution for Component Lifetimes | 工程学：部件寿命的指数分布

A production line produces LED bulbs whose lifetimes (in thousands of hours) are exponentially distributed with parameter λ. A quality engineer tests a random sample of 40 bulbs and records a total observed lifetime of 720 thousand hours. (a) Estimate λ and calculate the mean time to failure (MTTF). (b) Determine the probability that a bulb lasts longer than 25 000 hours. (c) Test the manufacturer’s claim that the true MTTF is at least 20 000 hours, using a 5% significance level.

某产线生产的 LED 灯泡寿命（千小时）服从参数为 λ 的指数分布。一位质量工程师随机抽取 40 只灯泡，记录总观测寿命为 720 千小时。(a) 估计 λ 并计算平均失效时间 MTTF；(b) 求一只灯泡寿命超过 25 000 小时的概率；(c) 检验制造商声称的“真实 MTTF 至少为 20 000 小时”，取 5% 显著性水平。

For an exponential distribution, the MTTF is 1/λ. The maximum likelihood estimator of λ is 1/x̄. Here x̄ = 720/40 = 18 thousand hours, so λ̂ = 1/18 ≈ 0.0556 per thousand hours. The estimated MTTF is 18 000 hours. The probability that a bulb lasts longer than 25 000 hours is P(X > 25) = e^−λ×25 = e^{−0.0556×25} ≈ e^−1.389 = 0.249. To test H₀: MTTF ≥ 20 (i.e. λ ≤ 0.05) against H₁: MTTF < 20 (λ > 0.05), we can use the sum of lifetimes T = ΣX_i which, under H₀ at the boundary λ = 0.05, follows a Gamma(40, 0.05) distribution, or equivalently 2λT ~ χ²₈₀. The observed total T = 720 gives χ² = 2×0.05×720 = 72. The lower‑tail p‑value for χ²₈₀ is about 0.257, which exceeds 0.05; we do not reject H₀. Thus, the data do not contradict the manufacturer’s claim. This problem integrates parameter estimation, probability calculation, and a hypothesis test shaped by the physical interpretation of lifetimes.

指数分布的 MTTF = 1/λ，λ 的极大似然估计为 1/x̄ = 1/18 ≈ 0.0556（每千小时），故估计 MTTF = 18 000 小时。P(X > 25) = e^{−0.0556×25} ≈ 0.249。检验 H₀: MTTF ≥ 20（λ ≤ 0.05）对 H₁: MTTF < 20 时，使用寿命总和 T = ΣX_i。在边界 λ = 0.05 下，T ~ Gamma(40, 0.05)，等效于 2λT ~ χ²₈₀。观测 T = 720 得 χ² = 72，其下尾 p 值约 0.257 > 0.05，不拒绝 H₀，数据未与制造商声明矛盾。该题融合了参数估计、概率计算和基于寿命物理意义的假设检验。

5. Medicine: Survival Analysis and Kaplan–Meier Estimation | 医学：生存分析与 Kaplan–Meier 估计

In a clinical trial, 12 patients with a serious disease are followed for 24 months. The recorded survival times (months) are: 6, 9, 12+, 15, 18, 20, 22, 24+, where ‘+’ indicates right‑censoring. (a) Construct the Kaplan–Meier estimate of the survival function. (b) Estimate the median survival time. (c) Draw a step‑function plot and explain how it handles censoring. (d) Outline the log‑rank test for comparing two such groups.

一项临床试验对 12 名重症患者随访 24 个月，记录生存时间（月）为：6, 9, 12+, 15, 18, 20, 22, 24+，其中 “+” 表示右删失。(a) 构建生存函数的 Kaplan–Meier 估计；(b) 估计中位生存时间；(c) 绘制阶梯函数图并解释如何处理删失；(d) 概述用于比较两组生存曲线的 log‑rank 检验。

We order the distinct event times: 6, 9, 15, 18, 20, 22 (ignoring censored times). At t = 0, Ŝ(0) = 1. At t = 6, there are 8 patients at risk (only 8 unique individuals? Actually total patients = 8 as listed, but check: list has 8 patients: 6,9,12+,15,18,20,22,24+. So n=8). One death at 6: Ŝ(6) = 1 × (7/8) = 0.875. At t = 9, at risk = 7, one death: Ŝ(9) = 0.875 × (6/7) = 0.750. The censored patient 12+ reduces the risk set but does not cause a drop. At t = 15, at risk = 5 (after removing 12+), one death: Ŝ(15) = 0.750 × (4/5) = 0.600. Continuing, Ŝ(18) = 0.600 × (3/4) = 0.450; Ŝ(20) = 0.450 × (2/3) = 0.300; Ŝ(22) = 0.300 × (1/2) = 0.150. The median survival time is the smallest t with Ŝ(t) ≤ 0.5, here 18 months. The Kaplan–Meier curve is a right‑continuous step function that drops only at observed event times; censored observations gracefully remove subjects from the risk set without creating a step. The log‑rank test compares observed and expected events under H₀ of equal hazards, producing a χ² statistic – essential for WJEC Pre‑U applied medicine questions.

将未删失事件时间排序：6, 9, 15, 18, 20, 22。初始生存概率为 1。t=6 时风险集 8 人，1 人去世，Ŝ(6)=0.875。t=9 时风险集 7 人，Ŝ(9)=0.750。删失患者 12+ 在风险集中移出但不降低曲线。后续 Ŝ(15)=0.600，Ŝ(18)=0.450，Ŝ(20)=0.300，Ŝ(22)=0.150。中位生存时间（Ŝ(t)≤0.5 的最小 t）为 18 个月。Kaplan–Meier 曲线是仅在死亡时间点跳跃的右连续阶梯函数，删失数据安静地缩减风险集。log‑rank 检验比较两组的观察与期望事件数，得出 χ² 统计量——是 Pre‑U 医学应用题中的常见考点。

6. Environmental Science: Chi‑Squared Goodness of Fit | 环境科学：拟合优度的卡方检验

An ecologist counts the number of a certain plant species in 100 randomly placed quadrats in a meadow. The frequencies of quadrats containing 0, 1, 2, 3, and ≥4 plants are 12, 28, 34, 18, and 8. Theory suggests that the distribution should be Poisson with mean 2.1. Test at the 5% significance level whether the data fit this Poisson model. Combine categories as necessary and comment on the degrees of freedom.

一位生态学家在草地上随机放置 100 个样方，统计某种植物在每样方中的数量，得到各频数：0 株 12 个，1 株 28 个，2 株 34 个，3 株 18 个，≥4 株 8 个。理论预测该分布应服从均值为 2.1 的泊松分布。在 5% 显著性水平下检验数据是否符合该泊松模型，必要时合并类别，并说明自由度。

First, compute expected frequencies under Poisson(2.1). P(X=0) = e⁻²·¹ = 0.1225, expected = 12.25; P(X=1) = 2.1×0.1225 = 0.2572, expected = 25.72; P(X=2) = 2.1²×0.1225/2 = 0.2701, expected = 27.01; P(X=3) = 2.1³×0.1225/6 = 0.1891, expected =

Published by TutorHao | Pre-U 统计 Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）