📚 A-Level Further Mathematics: F-Distribution Percentage Points | A-Level进阶数学：F分布百分位点
The F-distribution is central to statistical inference when comparing variances. In Edexcel A-Level Further Mathematics, you need to understand how to find and interpret its percentage points, often called critical values or quantiles. These values are used in hypothesis tests, particularly the F-test for equality of two population variances.
F分布在比较方差时的统计推断中处于核心地位。在Edexcel进阶数学（A-Level Further Mathematics）中，你需要理解如何求取并解释F分布的百分位点，这些点也常被称为临界值或分位数。它们用于假设检验，尤其是检验两个总体方差是否相等的F检验。
1. Definition and Origin of the F-Distribution | F分布的定义与来源
If \(U\) and \(V\) are independent random variables following chi-squared distributions with degrees of freedom \(\nu_1\) and \(\nu_2\) respectively, then the random variable
F = (U / ν₁) / (V / ν₂)
follows an F-distribution with parameters \(\nu_1\) (numerator degrees of freedom) and \(\nu_2\) (denominator degrees of freedom).
如果\(U\)和\(V\)分别是自由度为\(\nu_1\)和\(\nu_2\)的独立卡方分布随机变量，那么随机变量
F = (U / ν₁) / (V / ν₂)
服从以\(\nu_1\)（分子自由度）和\(\nu_2\)（分母自由度）为参数的F分布。
In practice, \(U\) and \(V\) typically arise from sample variances. For two independent samples from normal populations, the ratios of the biased or unbiased variance estimates produce an F statistic.
在实际应用中，\(U\)和\(V\)通常来自样本方差。对于来自正态总体的两个独立样本，样本方差估计值（有偏或无偏）之比构成F统计量。
2. Shape and Parameters of the F-Distribution | F分布的形态与参数
The F-distribution is right-skewed and takes only positive values. Its exact shape depends on the two degrees of freedom parameters, \(\nu_1\) and \(\nu_2\). Unlike the normal distribution, it is not symmetric, and its mean is approximately 1 when \(\nu_2\) is large.
F分布是右偏的，取值仅为正。其具体形态取决于两个自由度参数\(\nu_1\)和\(\nu_2\)。与正态分布不同，F分布不对称，当\(\nu_2\)较大时，其均值近似为1。
The probability density function is complex and is not required for calculation in Edexcel exams. However, you must be able to find percentage points using tables or calculator functions.
F分布的概率密度函数形式复杂，Edexcel考试不要求计算其具体值。但是，你必须能够通过查表或计算器函数求取百分位点。
3. What Is a Percentage Point? | 什么是百分位点？
For a continuous distribution, the percentage point \(x_p\) is the value such that \(P(F \le x_p) = p\). In hypothesis testing, we often need the upper-tail percentage point \(F_\alpha\), defined by \(P(F > F_\alpha) = \alpha\). This is equivalent to \(x_{1-\alpha}\).
对于连续分布，百分位点\(x_p\)是使得\(P(F \le x_p) = p\)的值。在假设检验中，我们通常需要上尾百分位点\(F_\alpha\)，定义为\(P(F > F_\alpha) = \alpha\)。这等价于\(x_{1-\alpha}\)。
For example, \(F_{0.05}\) is the value such that the probability to its right is 0.05. In an F-table, these upper-tail values are usually listed for common values of \(\alpha\) such as 0.10, 0.05, 0.025, and 0.01.
例如，\(F_{0.05}\)是右侧概率为0.05的值。在F表中，通常列出常见\(\alpha\)值（如0.10、0.05、0.025和0.01）对应的上尾值。
4. Structure of F-Distribution Tables | F分布表的结构
F-tables usually present critical values for a chosen upper-tail probability \(\alpha\). For each \(\alpha\), there is a separate table with rows corresponding to \(\nu_1\) and columns corresponding to \(\nu_2\).
F表通常针对选定的上尾概率\(\alpha\)给出临界值。对于每个\(\alpha\)，有单独的子表，其中行对应\(\nu_1\)，列对应\(\nu_2\)。
Suppose you need \(F_{0.05}(5, 10)\). You would locate the row for \(\nu_1 = 5\) and the column for \(\nu_2 = 10\) in the \(\alpha = 0.05\) table. The entry is approximately 3.33.
假设你需要\(F_{0.05}(5, 10)\)。你应在\(\alpha = 0.05\)子表中找到\(\nu_1 = 5\)所在的行与\(\nu_2 = 10\)所在的列，其交叉项约为3.33。
| α = 0.05 | ν₂ = 10 | ν₂ = 12 |
| ν₁ = 5 | 3.33 | 3.11 |
| ν₁ = 6 | 3.22 | 3.00 |
Always check whether the table gives upper-tail probabilities or cumulative probabilities. Edexcel formula booklets typically use upper-tail probabilities.
务必检查表中给出的是上尾概率还是累积概率。Edexcel公式书通常使用上尾概率。
5. Finding Percentage Points Using a Calculator | 用计算器求百分位点
Modern calculators approved for Edexcel exams, such as the Casio classwiz, provide an inverse F-distribution function. You can input the probability and both degrees of freedom directly.
Edexcel考试允许使用的现代计算器（如Casio ClassWiz）提供了F分布的逆累积分布函数。你可以直接输入概率和两个自由度。
For an upper-tail percentage point, if the calculator gives the inverse cumulative probability for \(P(F \le x)\), you must use \(1 – \alpha\) as the input. For example, to find \(F_{0.05}\), enter 0.95.
对于上尾百分位点，如果计算器给出的是累积概率\(P(F \le x)\)的逆函数，你必须输入\(1 – \alpha\)。例如，要求\(F_{0.05}\)，应输入0.95。
The exact menu names vary by model, but the principle is the same: locate the F inverse function, enter the probability, then the numerator and denominator degrees of freedom.
不同型号的计算器菜单名称可能不同，但原理相同：找到F分布的逆函数，输入概率，再输入分子和分母自由度。
6. The Reciprocal Property | 倒数性质
An important property of the F-distribution is reciprocal symmetry: if \(X \sim F(\nu_1, \nu_2)\), then \(1/X \sim F(\nu_2, \nu_1)\). This gives a useful formula for lower-tail critical values:
F分布有一个重要的倒数对称性质：如果\(X \sim F(\nu_1, \nu_2)\)，则\(1/X \sim F(\nu_2, \nu_1)\)。由此可得下尾临界值的有用公式：
Fα(ν₁, ν₂) = 1 / F1−α(ν₂, ν₁)
This means you do not need a separate lower-tail table. For example, to find \(F_{0.95}(5, 10)\), use the reciprocal property: \(F_{0.95}(5,10) = 1 / F_{0.05}(10,5)\).
这意味着你不需要单独的下尾表。例如，要求\(F_{0.95}(5, 10)\)，可用倒数性质：\(F_{0.95}(5,10) = 1 / F_{0.05}(10,5)\)。
7. Relationship with the t-Distribution and Chi-Squared Distribution | 与t分布和卡方分布的关系
If \(T \sim t(\nu)\), then \(T^2 \sim F(1, \nu)\). This relationship is used when testing whether a single coefficient is zero in regression analysis, though that is beyond the A-Level syllabuus. It also explains why a two-sided t-test can be performed as an F-test.
如果\(T \sim t(\nu)\)，则\(T^2 \sim F(1, \nu)\)。该关系用于回归分析中检验单个系数是否为零，尽管这超出了A-Level大纲。它同时也说明为何双侧t检验可以用F检验来完成。
Furthermore, if \(X \sim \chi^2(\nu)\), then \(X/\nu\) is a scaled version of the chi-squared distribution. Dividing two such independent scaled variables gives an F variable.
此外，如果\(X \sim \chi^2(\nu)\)，则\(X/\nu\)是卡方分布的缩放版本。两个独立的这类缩放变量相除即得F变量。
8. Application: Comparing Two Population Variances | 应用：比较两个总体方差
Suppose we have two independent random samples from normal populations with unknown variances \(\sigma_1^2\) and \(\sigma_2^2\). To test \(H_0: \sigma_1^2 = \sigma_2^2\), we use the statistic
假设我们有两个来自正态总体的独立随机样本，总体方差\(\sigma_1^2\)和\(\sigma_2^2\)未知。要检验\(H_0: \sigma_1^2 = \sigma_2^2\)，我们使用统计量
F = s₁² / s₂²
where \(s_1^2\) and \(s_2^2\) are the unbiased sample variances. Under \(H_0\), this statistic follows \(F(n_1-1, n_2-1)\).
其中\(s_1^2\)和\(s_2^2\)是无偏样本方差。在\(H_0\)下，该统计量服从\(F(n_1-1, n_2-1)\)分布。
For a one-tailed alternative \(H_1: \sigma_1^2 > \sigma_2^2\), we place \(s_1^2\) in the numerator and reject \(H_0\) if \(F > F_\alpha(\nu_1, \nu_2)\).
对于单尾备择假设\(H_1: \sigma_1^2 > \sigma_2^2\)，我们将\(s_1^2\)放在分子，若\(F > F_\alpha(\nu_1, \nu_2)\)则拒绝\(H_0\)。
9. Two-Tailed F-Test Critical Values | 双尾F检验的临界值
For a two-tailed test, we need both the lower and upper critical values. The upper critical value is \(F_{\alpha/2}(\nu_1, \nu_2)\). The lower critical value is found using the reciprocal property:
对于双尾检验，我们需要下临界值和上临界值。上临界值为\(F_{\alpha/2}(\nu_1, \nu_2)\)。下临界值通过倒数性质求得：
F1−α/2(ν₁, ν₂) = 1 / Fα/2(ν₂, ν₁)
For example, with \(\alpha = 0.05\), the lower tail probability is 0.025. To find the lower critical value for \(\nu_1 = 5, \nu_2 = 10\), we compute \(1 / F_{0.025}(10, 5)\).
例如，当\(\alpha = 0.05\)时，下尾概率为0.025。要求\(\nu_1 = 5, \nu_2 = 10\)时的下临界值，我们计算\(1 / F_{0.025}(10, 5)\)。
10. Common Pitfalls | 常见易错点
One frequent mistake is swapping the degrees of freedom. The numerator degrees of freedom always belong to the sample variance in the numerator of the F statistic.
一个常见错误是交换自由度。分子自由度始终属于F统计量分子中的样本方差。
Another error is using the wrong tail probability when using a calculator. Remember: inverse CDF calculators often output lower-tail cumulative probabilities, so for an upper-tail \(\alpha\) you must input \(1-\alpha\).
另一个错误是在使用计算器时用错了尾概率。记住：逆分布函数通常输出下尾累积概率，因此对于上尾\(\alpha\)，你必须输入\(1-\alpha\)。
Also, when using tables, verify whether the table is labelled by \(\alpha\) or by \(p\) (cumulative). A table labelled “0.95” for the upper tail actually gives \(F_{0.05}\). Read the heading carefully.
此外，查表时要确认表格标注的是\(\alpha\)还是\(p\)（累积）。如果上尾表标为“0.95”，实际上给出的是\(F_{0.05}\)。请仔细阅读表头。
11. Worked Example | 例题演示
Two independent samples of sizes \(n_1 = 12\) and \(n_2 = 9\) give sample variances \(s_1^2 = 45.6\) and \(s_2^2 = 18.2\). Test at the 5% significance level whether \(\sigma_1^2 > \sigma_2^2\).
两个独立样本的容量分别为\(n_1 = 12\)和\(n_2 = 9\)，样本方差为\(s_1^2 = 45.6\)和\(s_2^2 = 18.2\)。在5%显著性水平下检验\(\sigma_1^2 > \sigma_2^2\)是否成立。
Step 1: \(H_0: \sigma_1^2 = \sigma_2^2\), \(H_1: \sigma_1^2 > \sigma_2^2\).
第一步：\(H_0: \sigma_1^2 = \sigma_2^2\)，\(H_1: \sigma_1^2 > \sigma_2^2\)。
Step 2: \(F = 45.6 / 18.2 = 2.505\).
第二步：\(F = 45.6 / 18.2 = 2.505\)。
Step 3: Degrees of freedom are \(\nu_1 = 11\), \(\nu_2 = 8\). From tables, \(F_{0.05}(11, 8) = 3.31\). Since \(2.505 < 3.31\), we do not reject \(H_0\).
第三步：自由度为\(\nu_1 = 11\)，\(\nu_2 = 8\)。查表得\(F_{0.05}(11, 8) = 3.31\)。因为\(2.505 < 3.31\)，我们不拒绝\(H_0\)。
12. Summary | 总结
F-distribution percentage points are essential tools for variance-based hypothesis tests. You should be comfortable reading F-tables, using calculator inverse functions, applying the reciprocal property, and identifying correct degrees of freedom.
F分布百分位点是基于方差的假设检验的核心工具。你应该熟练查阅F表、使用计算器逆分布函数、灵活运用倒数性质，并准确识别自由度。
Mastering these skills will ensure you can tackle any F-distribution question in the Edexcel Further Statistics examination.
掌握这些技能，你就能从容应对Edexcel进阶数学统计考试中所有与F分布相关的问题。
Published by TutorHao | Further Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导