Voting Behaviour and the Media | 投票行为与媒体

📚 Voting Behaviour and the Media | 投票行为与媒体

In A-Level Mathematics, especially within the Edexcel Statistics modules, we often use real-world contexts to apply statistical techniques. The study of voting behaviour and media influence provides an excellent opportunity to explore descriptive statistics, correlation, regression, and hypothesis testing. By treating voter preferences and media exposure as variables, we can build mathematical models to investigate whether a relationship exists, and how strong it might be. This article will guide you through key statistical concepts, linking them directly to questions about voting and the media.

在A-Level数学中，特别是在Edexcel统计模块里，我们经常利用现实情境来应用统计方法。研究投票行为与媒体影响力，为探索描述性统计、相关性、回归分析和假设检验提供了绝佳的机会。通过将选民偏好和媒体接触视为变量，我们可以建立数学模型来探究是否存在某种关系，以及这种关系有多强。本文将带你梳理关键的统计学概念，并将它们直接与关于投票和媒体的问题联系起来。

---

1. Introduction to Voting Data | 投票数据简介

Election datasets typically contain a mix of categorical and numerical variables. A voter’s chosen party (e.g. Labour, Conservative, Liberal Democrat) is nominal data. Their age in years is continuous numerical data, while the number of hours spent watching political programmes on TV is discrete. Recognising these data types is the first step, as it determines which statistical diagrams and tests are appropriate. For example, we cannot calculate a meaningful mean for nominal data, but we can summarise it using frequencies and mode.

选举数据集通常包含类别变量和数值变量的混合。选民选择的政党（例如工党、保守党、自由民主党）属于名义数据。他们的年龄（以年为单位）是连续数值数据，而观看政治类电视节目的时数则是离散数据。识别这些数据类型是第一步，因为它决定了哪些统计图表和检验是合适的。例如，我们无法为名义数据计算有意义的均值，但可以用频数和众数来概括它。

Media exposure can be measured in several ways: hours of social media news consumption per day, number of political tweets seen, or a simple classification into ‘low’, ‘medium’, and ‘high’ usage. The latter is ordinal categorical data, allowing us to rank groups but not to quantify exact differences between them.

媒体曝光度可以通过多种方式衡量：每日阅读社交媒体新闻的小时数、看到的政治推文数量，或者简单分类为“低”“中”“高”使用度。后者是有序类别数据，允许我们对组别进行排序，但无法量化它们之间的确切差异。

---

2. Types of Data in Election Studies | 选举研究中的数据类型

When designing a survey on voting and media, you need to decide which variables to collect. Common variables include voting intention (nominal), trust in media (ordinal scale 1-5), weekly hours following politics (discrete), and income (continuous). In Edexcel S1 and S2, you learn that the choice between a chi-squared test, a t-test, or a correlation coefficient depends heavily on these variable types. For instance, to test whether media trust level is associated with voting intention, both being categorical, a chi-squared test for independence would be ideal.

在设计关于投票与媒体的调查时，你需要决定收集哪些变量。常见变量包括投票意向（名义）、对媒体的信任度（1-5等级有序）、每周关注政治的小时数（离散）和收入（连续）。在Edexcel S1和S2中你会学到，选择卡方检验、t检验或相关系数在很大程度上取决于这些变量类型。例如，要检验媒体信任度是否与投票意向相关，由于两者都是类别变量，使用独立性卡方检验是最理想的。

Numerical variables allow us to compute means, standard deviations, and correlation coefficients. Categorical variables lead us into frequency tables, bar charts, and chi-squared analysis. Understanding this division is vital for exam success.

数值变量允许我们计算均值、标准差和相关系数。类别变量则引导我们进入频数表、条形图和卡方分析。理解这一划分对于考试成功至关重要。

---

3. Summarising Voter Turnout | 概述选民投票率

Voter turnout, the percentage of eligible voters who actually vote, is a key numerical variable. We can calculate the mean turnout across different constituencies, the median, and the interquartile range to measure spread. If we suspect that media coverage affects turnout, we might compare the mean turnout in regions with high media coverage versus low coverage using a two-sample t-test (if conditions are met) or simply compare box plots to see if the differences are significant.

选民投票率，即实际投票的合格选民的百分比，是一个关键的数值变量。我们可以计算不同选区的平均投票率、中位数和四分位距来衡量离散程度。如果我们怀疑媒体报道会影响投票率，就可以比较媒体覆盖率高的地区与覆盖率低的地区的平均投票率，使用双样本t检验（如果条件满足），或者简单比较箱线图来看看差异是否显著。

For example, suppose we have the turnout figures for 30 constituencies where a local newspaper actively endorsed a candidate, and another 30 where it did not. The sample means and standard deviations can be used to construct a confidence interval for the difference in turnout, helping us infer whether the endorsement had an effect.

例如，假设我们获得了30个地方报纸积极为候选人代言的选区的投票率数据，以及另外30个没有代言的选区的数据。样本均值和标准差可用于构建投票率差异的置信区间，帮助我们推断代言是否产生了影响。

---

4. Measuring Media Exposure | 衡量媒体曝光度

To quantify media exposure, we might ask participants: ‘How many hours did you spend reading or watching political news in the past week?’ The collected data could look like the table below. A frequency distribution or histogram can reveal the shape of the distribution, whether it is symmetric or skewed. A box plot can quickly show outliers and the median. These descriptive tools help us summarise the central tendency and spread before comparing groups.

为了量化媒体曝光度，我们可能会问参与者：“在过去一周，您花了多少小时阅读或观看政治新闻？”收集到的数据可能如下表所示。频数分布或直方图可以揭示分布的形状，是对称还是偏态。箱线图可以快速显示异常值和中位数。这些描述性工具帮助我们在分组比较之前概括集中趋势和离散程度。

Hours of political news	Frequency
0 ≤ h < 2	12
2 ≤ h < 5	28
5 ≤ h < 10	35
10 ≤ h < 20	18
20 ≤ h < 30	7

From such grouped data, we can estimate the mean media hours and construct a cumulative frequency curve to find the median and percentiles. This process is an essential skill in A-Level Statistics.

从这样的分组数据中，我们可以估计平均媒体时数，并构建累积频数曲线来找到中位数和百分位数。这一过程是A-Level统计学中的一项基本技能。

---

5. Correlation vs. Causation | 相关性与因果关系

One of the most important principles in statistics is that correlation does not imply causation. We might observe a strong positive correlation between the number of hours spent watching a certain news channel and supporting a particular party. However, this could be because people who already support that party tend to watch that channel, not because the channel changes their minds. In an exam, you should always mention confounding variables and the possibility of reverse causation when interpreting a correlation coefficient.

统计中最重要的原则之一是相关性并不意味着因果关系。我们可能会观察到，观看某个新闻频道的小时数与支持某一特定政党之间存在很强的正相关。然而，这可能是因为那些本来就支持该政党的人往往收看那个频道，而不是因为这个频道改变了他们的想法。在考试中，解释相关系数时，你应该始终提及混杂变量和反向因果关系的可能性。

The Pearson product-moment correlation coefficient, r, measures the strength and direction of a linear relationship between two numerical variables. A value close to 1 indicates strong positive correlation, close to -1 strong negative, and near 0 no linear correlation. We can test whether r is significantly different from zero using a hypothesis test on the correlation coefficient.

皮尔逊积矩相关系数 r 衡量两个数值变量之间线性关系的强度和方向。数值接近1表示强正相关，接近-1表示强负相关，接近0表示没有线性相关。我们可以通过对相关系数进行假设检验，来检验 r 是否显著不同于零。

r = [ Σ (x – x̄)(y – ȳ) ] / √[ Σ (x – x̄)² Σ (y – ȳ)² ]

---

6. Scatter Plots and Regression Lines | 散点图与回归线

Before calculating r, always plot a scatter diagram. A scatter plot of ‘media hours per week’ against ‘support rating for Party A’ might reveal an outlier or a non-linear pattern. If the relationship appears roughly linear, we can fit a least squares regression line of the form y = a + bx. The slope b indicates how much y (support rating) changes per additional hour of media consumption. The coefficient b can be computed using b = Sxy / Sxx, where Sxy = Σ(x – x̄)(y – ȳ) and Sxx = Σ(x – x̄)².

在计算 r 之前，一定要先画散点图。以“每周媒体小时数”对“A政党的支持率”作散点图，可能会显示出异常值或非线性模式。如果关系看起来大致是线性的，我们就可以拟合一条形如 y = a + bx 的最小二乘回归线。斜率 b 表示每增加一小时媒体消费，y（支持率）改变了多少。系数 b 可以用 b = Sxy / Sxx 计算，其中 Sxy = Σ(x – x̄)(y – ȳ)，Sxx = Σ(x – x̄)²。

For example, imagine data from 50 towns. The regression line might be Support% = 22.5 + 3.2×(media hours). This suggests that each extra hour of political news per week is associated with a 3.2 percentage point increase in support, on average. But remember, this is an association, not necessarily an effect. The line can also be used to make predictions within the range of the data, but extrapolation should be avoided.

例如，假设有50个城镇的数据。回归线可能是支持率% = 22.5 + 3.2×(媒体小时数)。这表明每周每多收看一小时政治新闻，支持率平均增加3.2个百分点。但请记住，这是一种关联，不一定是影响。这条线还可以用来在数据范围内进行预测，但应避免外推。

---

7. Hypothesis Testing for Proportions | 比率的假设检验

Suppose a media analyst claims that after a televised debate, more than half of undecided voters now favour Candidate X. In a random sample of 200 undecided voters, 115 now favour Candidate X. We can test this claim using a one-tailed test on the population proportion p. The null hypothesis is H₀: p = 0.5, and the alternative is H₁: p > 0.5. Under H₀, the sample proportion p̂ follows an approximate normal distribution with mean 0.5 and standard error √[0.5×0.5/200] = 0.0354, provided np₀ > 5 and n(1 – p₀) > 5.

假设一位媒体分析师声称，在一次电视辩论之后，超过半数的未决定选民现在倾向于候选人X。在一个由200名未决定选民组成的随机样本中，有115人现在倾向于候选人X。我们可以使用对总体比例 p 的单尾检验来检验这一说法。原假设为 H₀: p = 0.5，备择假设为 H₁: p > 0.5。在原假设下，样本比例 p̂ 近似服从均值为0.5、标准误为 √[0.5×0.5/200] = 0.0354 的正态分布，只要 np₀ > 5 且 n(1 – p₀) > 5。

z = (p̂ – p) / √[p(1 – p)/n] = (0.575 – 0.5) / 0.0354 ≈ 2.12

At the 5% significance level, the critical z-value for a one-tailed test is 1.645. Since 2.12 > 1.645, we reject H₀ and conclude there is sufficient evidence that the debate increased support beyond 50%. A similar approach can be used to test whether the proportion of voters influenced by social media posts differs from a claimed value.

在5%的显著性水平下，单尾检验的临界z值是1.645。因为2.12 > 1.645，我们拒绝原假设，并得出结论：有足够的证据表明，辩论使支持率超过了50%。类似的方法可以用来检验受社交媒体帖子影响的选民比例是否与某个声称的值不同。

---

8. Chi-Squared Test for Independence | 独立性卡方检验

When both variables are categorical, the chi-squared test for independence is a powerful tool. Imagine we survey 300 people, recording their media diet (Predominantly TV, Social Media, or Print) and their voting intention (Party A, Party B, or Undecided). The data can be arranged in a 3×3 contingency table. The test determines whether there is a statistically significant association between media diet and voting intention.

当两个变量都是类别变量时，独立性卡方检验是一个强有力的工具。设想我们调查了300人，记录他们的媒体饮食（主要靠电视、社交媒体或纸质媒体）和投票意向（政党A、政党B或未决定）。这些数据可以排列成一个3×3的列联表。该检验能确定媒体饮食和投票意向之间是否存在统计学上显著的关联。

The expected frequency for each cell is calculated as (row total × column total) / grand total. The test statistic is χ² = Σ (O – E)² / E, where O is the observed frequency and E is the expected frequency. The degrees of freedom are (r – 1)(c – 1). For a 3×3 table, this is 4. If χ² exceeds the critical value from the chi-squared distribution at a chosen significance level (e.g. 9.488 for 4 df at 5%), we reject the null hypothesis of independence.

每个单元格的期望频数计算为（行总和 × 列总和）/ 总计。检验统计量为 χ² = Σ (O – E)² / E，其中 O 是观测频数，E 是期望频数。自由度为 (r-1)(c-1)。对于3×3表，自由度为4。如果 χ² 值超过了卡方分布在选定显著性水平下的临界值（例如，4个自由度5%水平下的9.488），我们就拒绝独立性的原假设。

χ² = Σ (O – E)² / E

This test does not tell us the strength or direction of the association, only that a relationship likely exists. In an exam, be sure to state the conclusion in context, such as ‘There is evidence to suggest that media type and voting intention are associated.’

这个检验并没有告诉我们关联的强度或方向，只是表明很可能存在某种关系。在考试中，一定要结合上下文陈述结论，例如“有证据表明媒体类型和投票意向之间存在关联”。

---

9. Interpreting p-values and Significance | 解释p值与显著性

The p-value is the probability of obtaining a test statistic at least as extreme as the one observed, assuming the null hypothesis is true. In the earlier proportion test, we could compute the exact p-value as P(Z > 2.12) = 0.017. Since 0.017 < 0.05, the result is significant at the 5% level. A small p-value indicates that the observed data would be very unlikely if H₀ were true, giving us reason to reject H₀.

p 值是在原假设成立的情况下，获得一个至少和观测值一样极端的检验统计量的概率。在先前的比例检验中，我们可以计算出确切的 p 值为 P(Z > 2.12) = 0.017。因为 0.017 < 0.05，所以在5%水平下结果是显著的。较小的 p 值表明，如果 H₀ 为真，观测到的数据会非常不可能出现，这给了我们拒绝 H₀ 的理由。

However, failing to reject H₀ does not prove H₀ is true; it simply means there is insufficient evidence against it. Media studies often suffer from small sample sizes, making it harder to achieve significance. Always report the p-value alongside the test statistic and state whether the result is significant at the given level.

然而，未能拒绝 H₀ 并不能证明 H₀ 为真；这只是意味着没有足够的证据反对它。媒体研究常受样本量小的困扰，这导致很难达到显著性。应当始终将 p 值与检验统计量一起报告，并说明在给定水平下结果是否显著。

---

10. Sampling Methods in Opinion Polls | 民意调查中的抽样方法

The validity of any statistical inference about voting and media depends heavily on the sampling method. A simple random sample gives every voter an equal chance of being selected, but it is often impractical. Stratified sampling divides the population into strata (e.g. age groups, regions) and takes random samples from each, ensuring representation. Quota sampling, often used in quick polls, sets quotas for

Published by TutorHao | A-Level Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）