📚 Common Misconceptions in SQA Statistics and How to Correct Them | SQA统计常见误区与纠正方法
Statistics is full of subtle traps that even confident students often fall into. For Year 11 learners following the SQA statistics curriculum, a handful of recurring mistakes appear again and again — mixing up types of average, misreading graphs, confusing correlation with causation, or mishandling probability. This article walks through ten of the most common misconceptions, explains exactly why they are wrong, and shows you how to avoid them. Each section pairs a clear English explanation with a matching Chinese version, helping you build solid understanding in both languages.
统计学中充满了微妙的陷阱，即使是很自信的学生也常常掉进去。对于正在学习 SQA 统计课程的 Year 11 学生来说，有几个反复出现的错误——混淆不同类型的平均数、误读图表、把相关关系当成因果关系，或者错误地处理概率。本文将带你梳理十个最常见的误区，清楚地解释为什么它们是错的，并告诉你如何避免。每个部分都配有清晰的英文解释和匹配的中文版本，帮助你在两种语言下建立扎实的理解。
1. Mixing Up Mean, Median, and Mode | 混淆平均数、中位数和众数
Many students treat 'average' as always meaning the mean, adding up all the values and dividing by the count. In datasets with outliers, the mean can be pulled far from the centre, making the median a better measure of typical value. The mode — the most frequent value — is often forgotten, yet it is the only average suitable for categorical data such as favourite colour or car brand.
很多学生一提到“平均数”就认为是把所有数值加起来再除以个数。在含有异常值的数据集中，均值会被拉得远离中心，此时中位数更能代表典型值。众数——出现次数最多的值——经常被遗忘，但它却是唯一适用于分类数据（如喜欢的颜色或汽车品牌）的平均数。
For example, in a salary dataset {£22k, £24k, £25k, £26k, £150k}, the mean is £49.4k, which does not reflect a typical salary. The median (£25k) does. When asked to find an average, first ask: 'Is there an outlier? What type of data is this?' Then choose the mean for symmetric numerical data, median for skewed data, and mode for categorical data.
例如，在新水数据集 {£22k, £24k, £25k, £26k, £150k} 中，均值是 49.4k，不能反映典型收入。中位数（25k）可以。当你被要求求平均数时，先问自己：“是否有异常值？这是什么类型的数据？”然后为对称数值数据选择均值，为偏斜数据选中位数，为分类数据选众数。
2. Confusing Correlation with Causation | 把相关关系误认为因果关系
Seeing two variables rise together does not mean one causes the other. A strong positive correlation between ice cream sales and drowning incidents does not imply ice cream causes drowning. Both are driven by a lurking variable: hot weather. The SQA exam often tests this distinction by presenting a high correlation coefficient and asking whether the relationship is causal.
看到两个变量一起上升，并不意味着一个导致了另一个。冰淇淋销量与溺水事件之间强烈的正相关，并不表示冰淇淋导致溺水。两者都受到一个潜在变量的影响：炎热的天气。SQA 考试经常通过呈现一个很高的相关系数，然后询问这种关系是否是因果关系，来考查这一区别。
rse causation. When interpreting r = 0.9 between hours of TV watched and test scores, ask: 'Could there be something else, like lack of sleep, that explains both?' Use phrases like 'is associated with' rather than 'causes' orrelation measures strength and direction of a linear relationship, not cause and effect.
要纠正这一错误，始终要考虑第三因素或反向因果关系。当解释看电视时间与考试分数之间的 r = 0.9 时，问自己：“是否有其他因素，比如睡眠不足，可以同时解释这两者？”使用“与……相关”等措辞，而不是“导致”。记住：相关性衡量线性关系的强度和方向，而不是因果关系。
3. Misreading Histograms: Frequency vs. Frequency Density | 误读直方图：频率与频率密度
A classic error is to assume the height of a histogram bar directly gives the frequency. In SQA statistics, when class widths are unequal, the vertical axis shows frequency density (frequency ÷ class width). A bar that is twice as tall may simply represent data from a narrower interval, not necessarily more observations.
一个经典错误是假设直方图条的高度直接给出频率。在 SQA 统计中，当组距不相等时，纵轴表示频率密度（频率 ÷ 组距）。一个高度是两倍的条，可能只是代表一个更窄的区间内的数据，而不一定是更多的观测数。
Always check the vertical axis label. If it says 'Frequency density' , a bar of density 4 over width 5 gives frequency 4 × 5 = 20. Drawing a histogram yourself? First calculate frequency densities, then plot them. This ensures area, not height, represents frequency.
始终检查纵轴标签。如果写的是“频率密度”，就用密度乘以组距来计算频率。例如，一个密度为 4、宽度为 5 的条，频率就是 4 × 5 = 20。自己画直方图？先计算频率密度，再绘制。这样才能确保面积而不是高度代表频率。
4. Believing a Larger Sample Guarantees Removal of Bias | 以为更大的样本能保证消除偏差
A widespread myth is that a big sample automatically produces reliable results. If the sampling method is biased — for example, only surveying people in a gym to estimate how many adults exercise regularly — increasing the sample size merely magnifies the bias. The estimate becomes more precise but remains systematically wrong.
一个普遍流传的误区是，大样本会自动产生可靠的结果。如果抽样方法有偏差——例如，只调查健身房的人来估计有多少成年人定期锻炼——增大样本量只会放大偏差。估计值会变得更精确，但仍然系统性地错误。
SQA questions often ask you to identify sampling issues. Always distinguish between random sampling error (which decreases with sample size) and sampling bias (which does not). To avoid bias, use simple random, stratified, or systematic sampling from an appropriate sampling frame. Check that every member of the population has an equal chance of being selected.
SQA 题目经常要求你识别抽样问题。始终区分随机抽样误差（会随样本量增大而减小）和抽样偏差（不会）。要避免偏差，应从合适的抽样框中采用简单随机、分层或系统抽样。确保总体中的每个成员都有相等的被选中机会。
5. Treating Probability as Guarantee in the Short Run | 把概率当作短期的保证
run relative frequency, not short-term certainty. Even with a fair coin, getting heads ten times in a row has a probability of (0.5)¹⁰ ≈ 0.001, which is unlikely but still possible.
在计算出某一天下雨的概率是 0.2 之后，学生经常预期在 10 天的假期里恰好有 2 天下雨。概率描述的是长期相对频率，而不是短期的确定性。即使是公平的硬币，连续 10 次正面朝上的概率也是 (0.5)¹⁰ ≈ 0.001，虽然不太可能但仍然可能发生。
0.2, the expected number of rainy days is 2, but the actual count could be 0, 1, 3, 4, etc., with non‑zero probabilities. Always say 'on average, we expect 2 rainy days' rather than 'there will be 2'.
更好的方法是，在承认变异性的前提下，用多次试验的期望值来思考。在一个 n=10, p=0.2 的二项分布情境中，下雨天数的期望值是 2，但实际天数可以是 0、1、3、4 等，其概率均不为零。始终说“平均而言，我们预计有 2 个雨天”，而不是“将会有 2 天”。
6. Conflating Statistical Significance with Practical Importance | 把统计显著性混同于实际重要性
When a hypothesis test yields a p-value below 0.05, students often declare the result 'important'. Statistical significance only means the observed effect is unlikely to have occurred by chance, given the sample size. A very large sample can make a tiny, trivial effect statistically significant.
当假设检验得出 p 值小于 0.05 时，学生常常宣称结果是“重要的”。统计显著性仅仅意味着，在给定的样本量下，观察到的效应不太可能偶然发生。一个非常大的样本可以让一个微小、无关紧要的效应变得具有统计显著性。
Always report the effect size alongside the p-value. For instance, a new teaching method might improve test scores by 0.1 marks on average. With thousands of students, p < 0.001, but is a 0.1‑mark gain worth the effort? The SQA expects you to interpret results in context, not just to rely on the p-value threshold.
始终要同时报告效应量和 p 值。例如，一种新的教学方法可能平均提高 0.1 分的考试成绩。在有数千名学生的情况下，p < 0.001，但 0.1 分的提高是否值得付出努力？SQA 期望你在上下文中解释结果，而不仅仅依赖 p 值阈值。
7. Confusing Sample Standard Deviation and Population Standard Deviation | 混淆样本标准差与总体标准差
A frequent computational error is using the wrong denominator in the standard deviation formula. The population standard deviation σ divides by n, while the sample standard deviation s divides by (n − 1). Using n for a sample underestimates variability, a mistake that SQA markers specifically watch for.
一个常见的计算错误是在标准差公式中使用了错误的分母。总体标准差 σ 除以 n，而样本标准差 s 除以 (n − 1)。对样本使用 n 会低估变异性，这是 SQA 阅卷人特别关注的错误。
When you have data from a sample and you want to estimate the population standard deviation, use s with (n − 1). Most calculator functions are labelled σn for population and σn₋₁ (or sx) for sample. Check your calculator mode and be explicit in your working: write the formula with Σ(x − x̄)²/(n − 1) and state 'since this is a sample, I divide by n − 1'.
当你持有样本数据并想估计总体标准差时，使用分母为 (n − 1) 的 s。大多数计算器功能标记为 σn 用于总体，σn₋₁（或 sx）用于样本。检查计算器模式，并在解题过程中明确写出其公式：Σ(x − x̄)²/(n − 1)，并说明“因为这是样本，我除以 n − 1”。
8. Misinterpreting Pie Charts and Bar Charts with Area Distortions | 误解饼图和存在面积扭曲的条形图
Pie charts should only be used when the categories sum to a meaningful whole. Mistake: using a pie chart for data where respondents could choose more than one option (multiple response), resulting in percentages that exceed 100%. Also, some bar charts deliberately start the y‑axis at a non‑zero value, exaggerating differences. Students often fail to notice this trick and draw misleading conclusions.
饼图仅应在分类数据总和构成一个有意义的整体时使用。误区：把饼图用于受访者可以选择多个选项（多选题）的数据，导致百分比之和超过 100%。此外，有些条形图故意将 y 轴起点设为非零值，从而夸大差异。学生经常注意不到这一把戏，得出误导性的结论。
When interpreting a chart, always read the axes, check the scale, and ask whether the visual impression matches the numbers. In a bar chart starting at 50 instead of 0, a bar of value 60 looks twice as tall as one of value 55, though the real difference is only 5. For SQA critical analysis questions, point out these distortions explicitly.
在解读图表时，始终要阅读坐标轴，检查刻度，并问自己视觉印象是否与数字相符。在一个起点为 50 而非 0 的条形图中，值为 60 的条看起来是值为 55 的条的两倍高，尽管实际差异只有 5。对于 SQA 的批判性分析题，要明确指出这些扭曲。
9. Misunderstanding the Regression Line: Prediction vs. Extrapolation | 误解回归直线：预测与外推
Students sometimes plug any x‑ + bx without checking whether it falls within the original data range. Predicting outside the range of x‑ absurd results because the linear relationship may not hold beyond the observed data. Yet in SQA tasks, this caution is frequently overlooked.
学生有时会把任何 x 值代入回归方程 y = a + bx，而不检查它是否在原数据范围内。在 x 值范围之外进行预测（外推）可能会产生荒谬的结果，因为线性关系在观测数据之外可能不再成立。然而在 SQA 任务中，这种谨慎经常被忽视。
For example, a regression of height (cm) on age (years) for children aged 5– = 80 + 5 × age. Plugging in age 40 would predict 280 cm, which is nonsense. Always state: 'This prediction is an extrapolation, so it is unreliable.' A safer approach is to restrict predictions to interpolation within the data range.
例如，对 5–12 岁儿童的身高（cm）关于年龄（岁）进行回归，可能得到身高 = 80 + 5 × 年龄。代入年龄 40 会预测出 280 厘米，这是荒谬的。始终声明：“这个预测是外推，因此不可靠。”更安全的方法是将预测限制在数据范围内的内插上。
10. Ignoring Assumptions in Hypothesis Tests and Confidence Intervals | 在假设检验和置信区间中忽视前提假设
Many tests taught in SQA, such as the t‑test or z‑test for a mean, assume the data come from a normal distribution or the sample size is large enough for the Central Limit Theorem to apply. Applying these tests to heavily skewed data with a small sample can lead to wrong conclusions. Students often skip checking normality or sample size requirements.
SQA 教授的许多检验，如均值的 t 检验或 z 检验，都假定数据来自正态分布，或者样本量足够大以使中心极限定理适用。将这些检验应用于小样本量的严重偏斜数据，可能导致错误结论。学生经常忽略检查正态性或样本量要求。
Before choosing a test, always state the assumptions and verify them if possible. For a small sample, use a boxplot or normal probability plot to assess symmetry and outliers. If the assumptions are violated, consider a non‑parametric alternative, or note the limitation in your conclusion. The phrase 'assuming the underlying population is approximately normal' is essential in SQA extended responses.
在选择检验方法之前，始终要陈述前提假设，并在可能的情况下进行验证。对于小样本，使用箱线图或正态概率图评估对称性和异常值。如果假设被违反，考虑使用非参数替代方法，或在结论中说明这一局限性。在 SQA 的扩展回答中，“假设总体近似正态”这一表述至关重要。
Published by TutorHao | Statistics 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