📚 Use of Large Sample Results for an Unknown Population | 利用大样本结果推断未知总体
In A-Level statistics, we often need to make inferences about a population parameter, such as the mean μ or proportion p, when the population distribution is unknown or its variance is unknown. When the sample size is sufficiently large, powerful approximate methods based on the Central Limit Theorem allow us to use sample statistics directly. This article explores how large-sample results are applied to construct confidence intervals and perform hypothesis tests for an unknown population mean and proportion, without needing to assume normality or know σ.
在A-Level统计中，我们经常需要对总体参数（如均值μ或比例p）进行推断，而总体分布可能是未知的或方差未知。当样本量足够大时，基于中心极限定理的强大近似方法允许我们直接使用样本统计量。本文将探讨如何利用大样本结果对未知总体均值和比例构建置信区间并进行假设检验，无需假定正态性或已知σ。
1. Central Limit Theorem Recap | 中心极限定理回顾
The Central Limit Theorem (CLT) states that for a random sample of size n drawn from any population with mean μ and finite variance σ², the sampling distribution of the sample mean x̄ becomes approximately normal with mean μ and standard error σ/√n, provided n is large enough. This approximation improves as n increases. This is the theoretical foundation for using large-sample methods.
中心极限定理指出，对于从任意均值为μ、方差为σ²（有限）的总体中抽取的样本量为n的随机样本，样本均值x̄的抽样分布近似服从均值为μ、标准误为σ/√n的正态分布，只要n足够大。随着n增大，近似程度改善。这是使用大样本方法的理论基础。
Even if the original population is heavily skewed or discrete, the distribution of x̄ tends to normality. Typically, n ≥ 30 is considered ‘arge’ for means, but this guideline may need to be larger for highly skewed distributions.
即使原始总体严重偏斜或离散，x̄的分布也趋向正态。通常，对于均值，n ≥ 30被视为“大样本”，但对于高度偏斜的分布，可能需要更大的样本量。
2. Unknown Population Variance: Using Sample Standard Deviation | 未知总体方差：使用样本标准差
When σ is unknown, we cannot compute the exact standard error σ/√n. However, if the sample size is large, we can replace σ with the sample standard deviation s, giving an estimated standard error s/√n. This substitution is justified because s is a consistent estimator of σ – as n grows, s converges to σ.
当σ未知时，我们无法计算精确的标准误σ/√n。然而，如果样本量很大，我们可以用样本标准差s代替σ，得到估计的标准误s/√n。这种替换是合理的，因为s是σ的一致估计量——随着n增大，s收敛到σ。
Consequently, the statistic (x̄ - μ) / (s/√n) follows an approximately standard normal distribution for large n, not a t-distribution. This allows us to use z-procedures instead of t-procedures when n is large.
因此，对于大n，统计量 (x̄ - μ) / (s/√n) 近似服从标准正态分布，而不是t分布。这使得我们在样本量较大时可以使用z方法而非t方法。
3. Confidence Intervals for the Mean (Large Sample, σ Unknown) | 均值的置信区间（大样本，σ未知）
A 100(1 - α)% confidence interval for the population mean μ when σ is unknown and n is large is given by:
x̄ ± zα/2 × (s / √n)
where zα/2 is the critical value from the standard normal distribution (e.g., 1.96 for 95% confidence).
当σ未知且样本量很大时，总体均值μ的100(1 - α)%置信区间为：
x̄ ± zα/2 × (s / √n)
其中zα/2是标准正态分布的临界值（例如，对于95%置信水平为1.96）。
Unlike the small-sample t-interval, this interval does not require the population to be normally distributed – the CLT ensures approximate normality of x̄. The interval's width depends on s, n, and the confidence level.
与小样本t区间不同，这个区间不要求总体正态分布——CLT保证了x̄的近似正态性。区间的宽度取决于s、n和置信水平。
Example: A sample of 50 students has mean score 72 and standard deviation 8. The 95% confidence interval for the true mean is 72 ± 1.96 × (8/√50) ≈ 72 ± 2.22, so (69.78, 74.22).
例子：一个包含50名学生的样本，平均分为72，标准差为8。真实均值的95%置信区间为 72 ± 1.96 × (8/√50) ≈ 72 ± 2.22，即 (69.78, 74.22)。
4. Hypothesis Tests for the Mean (Large Sample, σ Unknown) | 均值假设检验（大样本，σ未知）
To test H₀: μ = μ₀ vs H₁: μ ≠ μ₀ (or one-sided), we use the test statistic:
z = (x̄ - μ₀) / (s / √n)
Under H₀, this z follows an approximate N(0,1) distribution for large n. We compare z with critical values or use the p-value approach.
检验 H₀: μ = μ₀ 与 H₁: μ ≠ μ₀（或单侧），我们使用检验统计量：
z = (x̄ - μ₀) / (s / √n)
在H₀下，对于大n，该z近似服从N(0,1)分布。我们将z与临界值比较或使用p值法。
The rejection region for a two-tailed test at significance level α is |z| > zα/2 l Type I error rate may slightly differ from α, but the approximation is valid for n ≥ 30.
显著性水平α下的双侧检验拒绝域为 |z| > zα/2。由于使用了s，实际的第一类错误率可能与α略有差异，但对于n ≥ 30，近似是有效的。
excel A-Level, n ≥ 30 is the standard rule of thumb for using z-procedures with unknown σ.
始终检查背景：如果样本量小或总体极度非正态，使用非参数检验可能更安全。但在Edexcel A-Level中，n ≥ 30是使用含未知σ的z方法的标准经验法则。
5. Conditions and When n = 30 Works | 条件及n=30的适用性
The ‘ t works well for moderately skewed distributions. If the population is symmetric but heavy-tailed, a larger n might be needed for the normal approximation to be accurate in the tails. If there are outliers, the sample mean and standard deviation can be affected, so checking data is important.
“n ≥ 30”规则
Published by TutorHao | A-Level 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