描述统计:平均数是总和除以 n;中位数是排序后数据的中间值(n 为偶数时取中间两个数的平均)。样本标准差 s = √(Σ(x − x̄)²/(n − 1))。z 分数 z = (x − μ)/σ 表示某个值离平均数多少个标准差。
标准化
z = (66 − 60)/10 = 0.6
答案
z = 0.6
5.置信区间
答案: 上限 = 100.5
均值的置信区间:x̄ ± 误差范围。总体 σ 已知时,误差范围为 z·σ/√n;只有样本标准差 s 时,用自由度为 n − 1 的 t:t·s/√n。临界值:z = 1.645(90%)、1.96(95%)、2.576(99%)。要使误差范围不超过 E,需要 n ≥ (z·σ/E)²,并向上取整。
临界值
df = n − 1 = 20, t = 2.845
误差范围
E = 2.845·8/√21 = 4.967
区间
95.5 ± 4.967 → [90.53, 100.5]
答案
上限 = 100.5
6.置信区间
答案: n = 941
Confidence interval for a proportion: p̂ ± z·√(p̂(1 − p̂)/n), where p̂ = x/n. To reach a margin E you need n ≥ z²·p(1 − p)/E², rounded up; without a prior estimate use p = 0.5, which gives the largest n.
解出 n
n = 2.576²·0.15·0.85/0.03² = 940.1
向上取整
n = 941
答案
n = 941
7.假设检验与置信区间
答案: t = −3.398
Hypothesis test: the test statistic = (estimate − value under H₀)/(standard error under H₀). The p-value is the probability, assuming H₀ is true, of a statistic at least as extreme as the one observed: the area in one tail for a one-sided H₁, twice the tail area for a two-sided H₁. Reject H₀ when p ≤ α.
合并方差
sp² = (7·3.5² + 8·12²)/15 = 82.52
sp = 9.084
标准误
SE = 9.084·√(1/8 + 1/9) = 4.414
检验统计量
t = (120 − 135)/4.414 = −3.398
答案
t = −3.398
df = 15 时,双侧 5% 的临界值为 2.131;|t| 更大,所以拒绝 H₀。
8.假设检验与置信区间
答案: n = 482
检验的样本量:在水平 α 下的单侧 z 检验中,要以功效 1 − β 检测出真实偏移 δ,标准误必须满足 δ = (z_α + z_β)·σ/√n,所以 n = ((z_α + z_β)·σ/δ)²,向上取整。