📚 The F-distribution | F分布
The F-distribution is a continuous probability distribution that plays a central role in statistical inference, especially when comparing two population variances. Named after Sir Ronald Fisher, it forms the backbone of the analysis of variance (ANOVA) and many tests that examine ratios of variances. Understanding the F-distribution is essential for any A-Level Further Mathematics student tackling hypothesis tests on variances or constructing confidence intervals for variance ratios.
F分布是一种连续概率分布，在统计推断中扮演核心角色，尤其在比较两个总体方差时。它以罗纳德·费舍尔爵士的名字命名，是方差分析（ANOVA）的基础，也是许多检验方差比的方法的支柱。理解F分布对任何学习A-Level进阶数学、处理方差假设检验或构建方差比置信区间的学生都至关重要。
1. Definition and Construction | 定义与构造
The F-distribution is defined via two independent chi-squared random variables. If U ~ χ²(ν₁) and V ~ F-distribution with ν₁ numerator degrees of freedom and ν₂ denominator degrees of freedom.
F分布通过两个独立的卡方随机变量定义。如果U ~ χ²(ν₁) 和 V ~ χ²(ν₂) 相互独立，其中ν₁和ν₂为自由度，那么统计量 F = (U / ν₁) / (V / ν₂) 服从F分布，其分子自由度为ν₁，分母自由度为ν₂。
This construction explains why F appears naturally when we estimate the ratio of two population variances. For random samples from normal distributions, sample variances can be transformed into chi-squared variables, leading to an F-statistic.
这一构造解释了为什么在估计两个总体方差之比时F会自然出现。对于来自正态总体的随机样本，样本方差可以转化为卡方变量，从而得到F统计量。
2. Notation and Parameters | 记法与参数
We denote an F-distributed random variable as X ~ h the denominator. In applications, ν₁ = n₁ - 1 and ν₂ = n₂ - 1 when comparing variances of two independent samples of sizes n₁ and n₂.
我们记服从F分布的随机变量为 X ~ F(ν₁, ν₂)。参数的顺序很重要：ν₁始终是与分子卡方相关的自由度，ν₂与分母相关。在实际应用中，比较两个独立样本方差时，通常ν₁ = n₁ - 1，ν₂ = n₂ - 1，其中n₁和n₂为样本大小。
The support of the F-distribution is x > 0. It is skewed to the right and becomes more symmetric as both degrees of freedom increase.
F分布的支撑集为x > 0。它是右偏的，当两个自由度都增加时，分布变得更加对称。
3. Probability Density Function | 概率密度函数
The PDF of F(ν₁, ν₂) is given by:
F(ν₁, ν₂)的概率密度函数为：
f(x) = [Γ((ν₁+ν₂)/2) / (Γ(ν₁/2)Γ(ν₂/2))] · (ν₁/ν₂)ν₁/2 · xν₁/2 - 1 · (1 + (ν₁/ν₂)x)^(-(ν₁+ν₂)/2), x > 0
In the formula, Γ denotes the gamma function. While you are not required to memorise this complicated expression, it is useful to recognise the PDF's shape and the roles of ν₁ and ν₂.
公式中Γ表示伽马函数。虽然不需要记住这个复杂的表达式，但认识PDF的形状以及ν₁和ν₂的作用很有用。
The mean of F(ν₁, ν₂) is ν₂/(ν₂ - 2) for ν₂ > 2, and the variance is [2ν₂²(ν₁+ν₂-2)] / [ν₁(ν₂-2)²(ν₂-4)] for ν₂ > 4. These moments help characterise the distribution's central tendency and spread.
F(ν₁, ν₂)的均值为ν₂/(ν₂ - 2)（当ν₂ > 2时）；方差为[2ν₂²(ν₁+ν₂-2)] / [ν₁(ν₂-2)²(ν₂-4)]（当ν₂ > 4时）。这些矩特征有助于描绘分布的中心趋势和离散程度。
4. Key Properties | 关键性质
One important reciprocal property of the F-distribution is: if X ~ F(ν₁, ν₂), then 1/X ~ F(ν₂, ν₁). This relationship is extremely useful when using statistical tables, as most tables only provide upper-tail critical values.
F分布的一个重要的倒数性质是：如果X ~ F(ν₁, ν₂)，那么1/X ~ F(ν₂, ν₁)。这个关系在使用统计表时极为有用，因为大多数表格只提供上尾临界值。
The square of a t-distributed random variable with ν degrees of freedom follows an F-distribution: if T ~ t(ν), then T² ~ F(1, ν). This links hypothesis tests for a single mean and ANOVA for two groups.
自由度为ν的t分布随机变量的平方服从F分布：如果T ~ t(ν)，那么T² ~ F(1, ν)。这联系了单均值检验和两组方差分析的检验。
5. Using F-distribution Tables | F分布表的使用
Edexcel formulae booklets provide upper percentage points of the F-distribution for common significance levels (typically 5% and 1%). The table is organised with ν₁ (numerator) across the top and ν₂ (denominator) down the side. A typical entry is F(ν₁, ν₂; α), where α is the upper-tail probability.
Edexcel公式手册提供了常见显著性水平（通常为5%和1%）下F分布的上侧百分位数。表格的排列方式是：顶部为ν₁（分子自由度），侧面为ν₂（分母自由度）。典型的表值为F(ν₁, ν₂; α)，其中α为上尾概率。
-α) = 1 / F(ν₂, ν₁; α). Ensure you swap the degrees of freedom when taking the reciprocal.
要找出下尾临界值，可使用倒数性质：F(ν₁, ν₂; 1-α) = 1 / F(ν₂, ν₁; α)。取倒数时务必交换自由度。
6. The F-test for Equality of Two Variances | 双总体方差相等的F检验
² / s₂², where s₁² and s₂² are the unbiased sample variances from two independent normal samples.
最直接的应用之一是检验H₀: σ₁² = σ₂² 对 H₁: σ₁² ≠ σ₂² （双尾）或单侧备择假设。检验统计量为 F = s₁² / s₂²，其中s₁²和s₂²是来自两个独立正态样本的无偏样本方差。
To carry out a two-tailed test at significance level α, we compare F with upper and lower ½α critical values. The lower critical value is obtained using the reciprocal trick from tables.
在原假设下，该比值服从分子自由度为ν₁ = n₁ - 1、分母自由度为ν₂ = n₂ - 1的F分布。在显著性水平α下进行双尾检验时，我们将F与上下½α临界值比较。下临界值可利用表格的倒数技巧求得。
7. Confidence Interval for the Ratio of Two Variances | 两个方差比的置信区间
A 100(1-α)% confidence interval for σ₁²/σ₂² is given by:
σ₁²/σ₂² 的100(1-α)%置信区间为：
[ (s₁²/s₂²) × (1 / F(ν₁, ν₂; α/2)), (s₁²/s₂²) × (1 / F(ν₁, ν₂; 1-α/2)) ]
Again, the lower-tail F-critical value F(ν₁, ν₂; 1-α/2) can be found via 1 / F(ν₂, ν₁; α/2). The interval is not symmetric because the F-distribution is skewed.
同样，下尾F临界值F(ν₁, ν₂; 1-α/2)可以通过1 / F(ν₂, ν₁; α/2)求得。由于F分布是偏斜的，置信区间并不对称。
8. Assumptions for the F-test | F检验的假设条件
The F-test for variances relies on several assumptions: (1) the two samples are independent; (2) each sample is drawn from a normally distributed population; (3) the samples are random. The test is particularly sensitive to non-normality, so robust alternatives should be considered if normality is in doubt.
方差F检验依赖于几个假设：（1）两个样本相互独立；（2）每个样本来自正态分布总体；（3）样本是随机抽取的。该检验对非正态性特别敏感，因此如果对正态性存疑，应考虑稳健的替代方法。
In Edexcel problems, you will typically be told that populations are normally distributed or that the assumption may be assumed. Always state the assumptions and comment on their validity in context.
在Edexcel的题目中，通常会告知总体服从正态分布或可以假定正态。务必陈述假设并结合背景讨论其有效性。
9. Worked Example: Two-tailed Variance Test | 例题：双尾方差检验
nce s₁² = 1.44 g²; a sample of 13 from machine B has variance s₂² = 0.81 g². Test at the 5% level whether the population variances differ. Assume normality.
两台机器灌装瓶子。机器A的10瓶样本方差为s₁² = 1.44 g²；机器B的13瓶样本方差为s₂² = 0.81 g²。在5%显著性水平下检验总体方差是否不同。假定正态。
₁² = σ₂², H₁: σ₁² ≠ σ₂². ν₁ = 9, ν₂ = 12. Test statistic F = s₁²/s₂² = 1.44/0.81 ≈ 1.7778.
步骤1：H₀: σ₁² = σ₂², H₁: σ₁² ≠ σ₂²。ν₁ = 9, ν₂ = 12。检验统计量 F = s₁²/s₂² = 1.44/0.81 ≈ 1.7778。
Step 2: Upper 2.5% 975) = 1 / F(12,9; 0.025) ≈ 1/3.87 ≈ 0.258. The rejection region is F < 0.258 or F > 3.07.
步骤2：从表中查得上2.5%临界值：F(9,12; 0.025) ≈ 3.07（根据提供表格）。下临界值：F(9,12; 0.975) = 1 / F(12,9; 0.025) ≈ 1/3.87 ≈ 0.258。拒绝域为F < 0.258 或 F > 3.07。
Step 3: Since 0.258 < 1.7778 < 3.07, we do not reject H₀. There is insufficient evidence of a difference in variances.
步骤3：因为0.258 < 1.7778 < 3.07，不拒绝H₀。没有足够证据表明方差存在差异。
10. Connection with Analysis of Variance (ANOVA) | 与方差分析的联系
Although a full treatment of ANOVA is beyond the scope of Edexcel FS1, the F-distribution underpins one-way ANOVA. The F-statistic compares the variance between sample means to the variance within samples, and follows an F-distribution under the null hypothesis that all group means are equal.
虽然完整的方差分析不在Edexcel FS1的范围内，但F分布奠定了单因素方差分析的基础。F统计量比较了样本均值之间的方差和样本内部的方差，并在所有组均值相等的原假设下服从F分布。
F-values indicate that between-group variability is large relative to within-group variability, leading to rejection of H₀.
分子自由度 = k - 1（k为组数），分母自由度 = N - k（N为总观测数）。F值大表明组间变异相对于组内变异较大，从而拒绝原假设。
11. Common Mistakes and Tips | 常见错误与提示
other pitfall is using the wrong order for the ratio F = s₁²/s₂²; be consistent with the hypotheses. For a two-tailed test, remember to halve the significance level for each tail.
学生们常常在查找下尾临界值时忘记交换自由度。务必检查：下侧F(ν₁, ν₂) = 1 / 上侧F(ν₂, ν₁)。另一个陷阱是比值F = s₁²/s₂²的顺序错误；确保与假设一致。对于双尾检验，记得每个尾部使用显著性水平的一半。
Also, never assume the interval for a variance ratio is symmetric; use the correct upper and lower critical values from the F-tables. Check that the larger variance is not automatically placed in the numerator without justification, as this can affect the test decision.
此外，切勿假定方差比的区间是对称的；应使用F表中正确的上下临界值。不要未经论证就自动将较大的方差放在分子中，因为这会影响检验决策。
12. Summary and Revision Checklist | 总结与复习清单
The F-distribution is a flexible tool for inference about population variances. Key points: defined via independent chi-squared variables; parameters ν₁, ν₂ are numerator and denominator df; reciprocal property simplifies lower-tail critical values; used in F-test for equal variances and confidence intervals for σ₁²/σ₂²; sensitive to normality assumptions. Ensure you can use the Edexcel statistical tables confidently to extract upper percentage points and apply the reciprocal property.
F分布是推断总体方差的灵活工具。要点：通过独立的卡方变量定义；参数ν₁, ν₂为分子和分母自由度；倒数性质简化了下尾临界值；用于方差相等的F检验和σ₁²/σ₂²的置信区间；对正态性假设敏感。务必能够熟练使用Edexcel统计表提取上侧百分位数并应用倒数性质。
A solid grasp of the F-distribution not only prepares you for the FS1 examination but also provides a strong foundation for further study in statistics.
扎实掌握F分布不仅能为FS1考试做好准备，也为进一步的统计学学习奠定坚实基础。
Published by TutorHao | Further 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