📚 Application of the F-Test | F检验的应用
The F-test is a cornerstone of statistical inference in A-Level Further Mathematics, particularly for comparing variances and analysing variance in experimental designs. It is built upon the F-distribution, first developed by Sir Ronald Fisher, and serves as a powerful tool for hypothesis testing when dealing with normally distributed data.
F检验是A-Level进阶数学中统计推断的核心内容，尤其在比较方差和分析实验设计中的方差时应用广泛。它基于由罗纳德·费希尔爵士首创的F分布，是处理正态分布数据时进行假设检验的强有力工具。
1. Understanding the F-Distribution | 理解F分布
nt chi-squared random variables, each divided by its respective degrees of freedom. If X₁² ~ χ²(m) and X₂² ~ ²/m) ÷ (X₂²/n) follows an F-distribution with m and n degrees of freedom, denoted F(m, n).
F分布是一种连续概率分布，当计算两个独立卡方随机变量分别除以其自由度后的比值时产生。若 X₁² ~ χ²(m) 与 X₂² ~ χ²(n) 相互独立，则统计量 F = (X₁²/m) ÷ (X₂²/n) 服从自由度为 m 和 n 的F分布，记作 F(m, n)。
Key properties of the F-distribution include its positive skewness, which decreases as the degrees of freedom increase, and its mean of approximately 1 for large denominator degrees of freedom. Since F ratios are always positive, the distribution is defined only for x > 0.
F分布的主要性质包括其正偏态，且偏度随自由度的增加而减小；当分母自由度较大时，均值接近1。由于F比值始终为正，分布仅在 x > 0 时有定义。
F = (χ₁² / m) ÷ (χ₂² / n) = (S₁² / m) × (n / S₂²)
In practice, for two independent samples of sizes n₁ and n₂ taken from normal populations, the statistic:
在实际应用中，对于来自正态总体的两个独立样本，其容量分别为 n₁ 和 n₂，统计量定义为：
F = s₁² / s₂² ~ F(n₁ − 1, n₂ − 1)
les, the larger sample variance is conventionally placed in the numerator, ensuring F ≥ 1.
其中 s₁² 和 s₂² 分别是两个样本的方差。为方便使用标准F分布表，通常将较大的样本方差放在分子位置，以确保 F ≥ 1。
2. One-Tailed vs Two-Tailed F-Tests | 单尾与双尾F检验
ve is H₁: σ₁² > σ₂². The critical region lies entirely in the upper tail of the F-distribution.
当检验一个总体方差是否大于另一个时，使用单尾检验。原假设为 H₀: σ₁² = σ₂²，备择假设为 H₁: σ₁² > σ₂²。拒绝域完全位于F分布的上尾。
For a two-tailed test, where H₁: σ₁² ≠ σ₂², we use both tails. However, by convention, we always place the larger sample variance in the numerator and use the upper-tail critical value F(α/2; n₁ − 1, n₂ − 1), because the lower tail can be expressed as 1 / F(α/2; n₂ − 1, n₁ − 1).
对于双尾检验，备择假设为 H₁: σ₁² ≠ σ₂²，需要使用两个尾部。然而按照惯例，我们总是将较大的样本方差放在分子位置，并使用上尾临界值 F(α/2; n₁ − 1, n₂ − 1)，因为下尾可以表示为 1 / F(α/2; n₂ − 1, n₁ − 1)。
In Edexcel A-Level Further Mathematics, the majority of F-test questions are one-tailed, as they compare a claimed variance against a sample variance, or two sample variances where prior evidence suggests a direction.
在Edexcel A-Level进阶数学考试中，大多数F检验题目为单尾检验，例如将一个声称的方差与样本方差比较，或者在已有证据提示方向时比较两个样本方差。
3. Critical Values and F-Tables | 临界值与F分布表
ach table corresponds to a specific upper-tail probability, with rows representing numerator degrees of freedom (ν₁) and columns representing denominator degrees of freedom (ν₂).
F分布表提供了不同显著性水平下的临界值，常见的有 α = 0.10、0.05、0.01 和 0.025。每个表格对应特定的上尾概率，其中行代表分子自由度 (ν₁)，列代表分母自由度 (ν₂)。
in the table for the desired significance level. For example, the critical value at the 5% significance level with ν₁ = 5 and ν₂ = 10 is F(0.05; 5, 10) = 3.33.
要查找临界值 F(α; ν₁, ν₂)，请在对应显著性水平的表格中找到第 ν₁ 行与第 ν₂ 列的交叉点。例如，显著性水平为5%时，自由度为 ν₁ = 5 和 ν₂ = 10 的临界值为 F(0.05; 5, 10) = 3.33。
It is crucial to note that F-tables are one-tailed. For a two-tailed test at significance level α, we use the table for α/2. Many candidates lose marks by using the wrong column or misreading the degrees of freedom.
必须注意，F分布表是单尾的。对于显著性水平为 α 的双尾检验，应使用 α/2 的表格。许多考生因使用错误的列或误读自由度而失分。
4. Comparing Two Population Variances | 比较两个总体方差
ons) and s₂² (with n₂ observations), the test statistic is F = s₁² / s₂², where s₁² ≥ s₂².
F检验最常见的应用是比较两个独立正态分布总体的方差。已知样本方差 s₁²（包含 n₁ 个观测值）和 s₂²（包含 n₂ 个观测值），检验统计量为 F = s₁² / s₂²，其中 s₁² ≥ s₂²。
| Step | Action | 步骤 | 操作 |
| 1 | State hypotheses H₀: σ₁² = σ₂² and H₁: σ₁² > σ₂² (or ≠) | 1 | 写出原假设 H₀: σ₁² = σ₂² 与备择假设 H₁: σ₁² > σ₂²（或 ≠） |
| 2 | Calculate F = s₁² / s₂² with s₁² ≥ s₂² | 2 | 计算 F = s₁² / s₂²，其中 s₁² ≥ s₂² |
| 3 | Determine degrees of freedom ν₁ = n₁ − 1, ν₂ = n₂ − 1 | 3 | 确定自由度 ν₁ = n₁ − 1, ν₂ = n₂ − 1 |
| 4 | Find critical value from F-tables | 4 | 从F分布表中查找临界值 |
| 5 | Compare and conclude | 5 | 比较并得出结论 |
m line A (n₁ = 12) give s₁² = 4.8, and from line B (n₂ = 10) give s₂² = 2.1. Test at the 5% level whether the variance of line A is significantly greater than that of line B.
某制造商声称两条生产线生产的部件具有相同的变异性。从A线抽取样本 (n₁ = 12) 得到 s₁² = 4.8，从B线抽取样本 (n₂ = 10) 得到 s₂² = 2.1。在5%显著性水平下检验A线的方差是否显著大于B线。
H₀: σ₁² = σ₂², H₁: σ₁² > σ₂²; F = 4.8 / 2.1 = 2.286
.10. Since 2.286 < 3.10, we do not reject H₀. There is insufficient evidence to conclude that the variances differ significantly.
自由度为 ν₁ = 11 和 ν₂ = 9。临界值 F(0.05; 11, 9) = 3.10。由于 2.286 < 3.10，我们不能拒绝原假设 H₀。没有足够证据表明方差存在显著差异。
5. F-Test in One-Way ANOVA | 单因素方差分析中的F检验
The F-test is also central to one-way Analysis of Variance (ANOVA), where we compare the means of three or more independent groups. The total variation in the data is partitioned into between-group variation (SSB) and within-group variation (SSW).
F检验同样是单因素方差分析（ANOVA）的核心，在该分析中我们比较三个或更多独立组的均值。数据的总变异被分解为组间变异 (SSB) 和组内变异 (SSW)。
The ANOVA F-statistic is the ratio of the mean square between groups (MSB) to the mean square within groups (MSW):
方差分析的F统计量是组间均方 (MSB) 与组内均方 (MSW) 之比：
F = MSB / MSW = [SSB / (k − 1)] ÷ [SSW / (N − k)]
Here, k is the number of groups and N is the total number of observations. Under H₀: all group means are equal, F follows an F-distribution with (k − 1) and (N − k) degrees of freedom.
其中 k 是组数，N 是总观测数。在原假设 H₀：所有组均值相等 下，F 服从自由度为 (k − 1) 和 (N − k) 的F分布。
In the Edexcel specification, you are expected to be able to complete an ANOVA table given summary statistics, or to calculate the test statistic from raw data using the relevant sums of squares.
在Edexcel考试大纲中，要求考生能够根据汇总统计量完成方差分析表，或利用原始数据计算相关平方和从而得到检验统计量。
6. The ANOVA Table | 方差分析表
The standard ANOVA table organises the calculations systematically. The explained and unexplained variation are separated, allowing the F-statistic to be evaluated efficiently.
标准方差分析表系统性地组织计算过程。它分离了解释变异和未解释变异，从而可以高效地评估F统计量。
| Source of Variation | Sum of Squares | df | Mean Square | F |
| Between groups | SSB | k − 1 | MSB = SSB / (k − 1) | MSB / MSW |
| Within groups | SSW | N − k | MSW = SSW / (N − k) | — |
| Total | TSS | N − 1 | — | — |
Where TSS = SSB + SSW. The total sum of squares is calculated as Σ(xᵢⱼ − x̄)², the between-group sum of squares as Σ nᵢ(x̄ᵢ − x̄)², and the within-group sum of squares by subtraction.
其中 TSS = SSB + SSW。总平方和按 Σ(xᵢⱼ − x̄)² 计算，组间平方和按 Σ nᵢ(x̄ᵢ − x̄)² 计算，组内平方和则通过减法获得。
1 − 71)² = 45 + 45 + 0 = 90. MSB = 90/2 = 45, MSW = 210/12 = 17.5, and F = 45/17.5 = 2.57.
考虑对15名学生（每组5人）实施三种教学方法。各组均值分别为68、74和71，总均值为71，且 SSW = 210。则 SSB = 5(68 − 71)² + 5(74 − 71)² + 5(71 − 71)² = 45 + 45 + 0 = 90。MSB = 90/2 = 45，MSW = 210/12 = 17.5，因此 F = 45/17.5 = 2.57。
At the 5% 7 < 3.89, there is insufficient evidence to reject the null hypothesis; the teaching methods do not produce significantly different mean scores.
在5%显著性水平下，临界值为 F(0.05; 2, 12) = 3.89。由于 2.57 < 3.89，没有足够证据拒绝原假设；三种教学方法产生的平均分没有显著差异。
7. Assumptions of the F-Test | F检验的假设条件
For the F-test to be valid, several assumptions must be satisfied. First, the populations from which samples are drawn must be approximately normally distributed. The F-test is sensitive to departures from normality, especially when sample sizes are small.
为保证F检验的有效性，必须满足若干假设条件。首先，样本所来自的总体应近似服从正态分布。F检验对正态性的偏离较为敏感，尤其当样本量较小时。
Second, the samples must be independent and randomly selected. This ensures that the observed variation accurately reflects population characteristics rather than sampling artefacts. For two-sample variance tests, the two samples must be independent of each other.
其次，样本必须独立且随机选取。这确保观察到的变异能够真实反映总体特征，而非抽样伪迹。对于双样本方差检验，两个样本必须相互独立。
Third, in ANOVA, the population variances must be equal (homoscedasticity). This is the very assumption being tested when we apply the F-test to variances, and in ANOVA it is a requirement for the F-test on means to be reliable.
第三，在方差分析中，各总体方差必须相等（方差齐性）。当我们将F检验应用于方差时，这恰恰是所检验的假设；而在方差分析中，方差齐性也是均值F检验可靠的前提。
Candidates should always state these assumptions when carrying out a hypothesis test in an exam setting, as it demonstrates a full understanding of the procedure.
在考试中进行假设检验时，考生应始终说明这些假设条件，这能体现对检验流程的完整理解。
8. Relationship Between F and t Distributions | F分布与t分布的关系
A beautiful theoretical result links the F and t distributions: if a random variable T follows a t-distribution with ν degrees of freedom, then T² follows an F-distribution with 1 and ν degrees of freedom. That is, T² ~ F(1, ν).
一条优美的理论结果将F分布与t分布联系起来：若随机变量 T 服从自由度为 ν 的t分布，则 T² 服从自由度为 1 和 ν 的F分布，即 T² ~ F(1, ν)。
This relationship explains why testing the equality of two variances yields the same conclusion as a two-sample t-test for means only when the latter assumes equal variances. It also explains why the square of a t-statistic equals the corresponding F-statistic in certain regression contexts.
这一关系解释了为什么在t检验假设方差相等的前提下，比较两个方差与双样本均值t检验会得出相同的结论；也解释了在某些回归分析中，t统计量的平方恰好等于相应的F统计量。
If T ~ t(ν), then T² ~ F(1, ν)
This result is occasionally tested in Edexcel examinations as a short theoretical question, requiring candidates to recall or derive the degrees-of-freedom relationship.
这一结果偶尔会作为简答题出现在Edexcel考试中，要求考生回忆或推导自由度的关系。
9. Worked Exam-Style Example | 考试风格例题精解
A machine produces precision rods. The manufacturer claims that the standard deviation of rod lengths is no more than 2.0 mm. A random sample of 16 rods yields a sample standard deviation of 2.8 mm. Test at the 1% significance level whether the true standard deviation exceeds 2.0 mm.
某机器生产精密杆件。制造商声称杆件长度的标准差不超过 2.0 mm。随机抽取16根杆件，样本标准差为 2.8 mm。在1%显著性水平下检验真实标准差是否超过 2.0 mm。
e ν₁ = n − 1 = 15 and ν₂ = ∞ (since the claimed variance is a fixed population parameter).
此处检验单一方差是否大于声称值。检验统计量为 F = s² / σ₀² = 2.8² / 2.0² = 7.84 / 4.00 = 1.96。自由度为 ν₁ = n − 1 = 15，ν₂ = ∞（因为声称的方差是固定的总体参数）。
/ σ₀² = 15 × 7.84 / 4.00 = 29.4. This is the more appropriate test for a single variance.
然而，标准F分布表通常不包含 ν₂ = ∞ 的情况。这种情况下，我们改用卡方分布：χ² = (n − 1)s² / σ₀² = 15 × 7.84 / 4.00 = 29.4。这才是检验单一方差的更合适方法。
With 15 degrees of freedom, the upper critical value at the 1% 29.4 < 30.578, we fail to reject H₀. There is insufficient evidence to conclude that the standard deviation exceeds 2.0 mm.
自由度为15时，1%水平的上尾临界值为 χ²(0.01; 15) = 30.578。由于 29.4 < 30.578，我们不能拒绝原假设。没有足够证据表明标准差超过 2.0 mm。
This example illustrates a crucial point: the F-test for two variances requires two samples. When comparing a sample variance to a known population variance, the chi-squared test is the correct procedure.
此例说明了一个关键点：双样本方差比较使用F检验；而当将样本方差与已知总体方差比较时，应使用卡方检验。
10. Common Pitfalls and Exam Tips | 常见错误与考试技巧
el if performing a two-tailed test. Misidentifying the degrees of freedom is another common pitfall: for two samples of sizes n₁ and n₂, the degrees of freedom are (n₁ − 1, n₂ − 1), not (n₁, n₂).
一个常见错误是将较小方差放在分子位置。务必安排检验统计量使 F ≥ 1，若进行双尾检验还需将显著性水平减半。另一个常见陷阱是误判自由度：对于容量为 n₁ 和 n₂ 的两个样本，自由度应为 (n₁ − 1, n₂ − 1)，而非 (n₁, n₂)。
When using a calculator, ensure you have selected the correct distribution function. Many calculators include an inverse F function; double-check that the inputs correspond to the correct numerator and denominator degrees of freedom.
使用计算器时，务必选择正确的分布函数。许多计算器内置F分布的逆函数；请仔细检查输入值是否对应正确的分子和分母自由度。
In conclusion, the F-test is a versatile tool for comparing variances and conducting ANOVA. Mastering its mechanics, understanding the underlying assumptions, and practising with past paper questions are essential for success in Edexcel A-Level Further Mathematics.
总而言之，F检验是比较方差和进行方差分析的多功能工具。掌握其计算步骤、理解基本假设条件，并通过历年真题勤加练习，是在Edexcel A-Level进阶数学中取得优异成绩的关键。
Always state your hypotheses in terms of population parameters, show the test statistic clearly, and write a concluding sentence that answers the original question in context.
始终以总体参数的形式写出原假设和备择假设，清晰展示检验统计量的计算过程，并写出结合题目背景的结论句。
Published by TutorHao | Further Mathematics 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