X 1 , ⋯ , X n i.i.d. , ∼ N ( 0 , 1 ) Y = ∑ i = 1 n X i 2 X_{1},\cdots,X_{n}\; \textup{i.i.d.},\sim N(0,1)\\Y=\sum_{i=1}^{n}X_{i}^{2} X1,⋯,Xni.i.d.,∼N(0,1)Y=i=1∑nXi2 Y Y Y服从自由度为 n n n的卡方分布: Y ∼ χ 2 ( n ) Y\sim \chi^{2}(n) Y∼χ2(n)重要性质: X 1 ∼ χ 2 ( m ) , X 2 ∼ χ 2 ( n ) , X 1 , X 2 独立 ⇒ X 1 + X 2 ∼ χ 2 ( m + n ) \boxed{X_{1}\sim \chi^{2}(m),\;X_{2}\sim \chi^{2}(n),\;X_{1},X_{2}独立\;\Rightarrow\; X_{1}+X_{2}\sim \chi^{2}(m+n)} X1∼χ2(m),X2∼χ2(n),X1,X2独立⇒X1+X2∼χ2(m+n) X 1 , ⋯ , X n i.i.d. , ∼ E ( λ ) ⇒ 2 λ ∑ i = 1 n X i ∼ χ 2 ( 2 n ) X_{1},\cdots,X_{n}\;\textup{i.i.d.},\sim E(\lambda)\;\Rightarrow\; 2\lambda\sum_{i=1}^{n}X_{i}\sim \chi^{2}(2n) X1,⋯,Xni.i.d.,∼E(λ)⇒2λi=1∑nXi∼χ2(2n)
X 1 ∼ χ 2 ( n ) , X 2 ∼ N ( 0 , 1 ) , X 1 , X 2 独立 X_{1}\sim \chi^{2}(n),\; X_{2}\sim N(0,1),\; X_{1},X_{2}独立 X1∼χ2(n),X2∼N(0,1),X1,X2独立 Y = X 2 1 n X 1 Y=\frac{X_{2}}{\sqrt{\frac{1}{n}X_{1}}} Y=n1X1X2 Y Y Y服从自由度为 n n n的 t t t分布: Y ∼ t ( n ) Y\sim t(n) Y∼t(n)重要性质: 当 n → ∞ 时 , t 分布 t ( n ) 收敛于标准正态分布 N ( 0 , 1 ) . 一般认为当 n > 30 ( 一说 n > 45 ) 时 , 有 Y = X 2 1 n X 1 ∼ 近似 N ( 0 , 1 ) . \boxed{当n\to \infty时,\; t分布t(n)收敛于标准正态分布N(0,1).\;一般认为当n>30\,(一说n>45)\,时,\; 有Y=\frac{X_{2}}{\sqrt{\frac{1}{n}X_{1}}}\overset{近似}{\sim}N(0,1).} 当n→∞时,t分布t(n)收敛于标准正态分布N(0,1).一般认为当n>30(一说n>45)时,有Y=n1X1X2∼近似N(0,1).
X 1 ∼ χ 2 ( n ) , X 2 ∼ χ 2 ( m ) , X 1 , X 2 独立 X_{1}\sim \chi^{2}(n),\; X_{2}\sim \chi^{2}(m),\; X_{1},X_{2}独立 X1∼χ2(n),X2∼χ2(m),X1,X2独立 Y = 1 m X 2 1 n X 1 Y=\frac{\frac{1}{m}X_{2}}{\frac{1}{n}X_{1}} Y=n1X1m1X2 Y Y Y服从自由度为 ( m , n ) (m,n) (m,n)的 F F F分布: Y ∼ F ( m , n ) Y\sim F(m,n) Y∼F(m,n)重要性质: X ∼ F ( m , n ) ⇒ 1 X ∼ F ( n , m ) \boxed{X\sim F(m,n)\; \Rightarrow\; \frac{1}{X}\sim F(n,m)} X∼F(m,n)⇒X1∼F(n,m) X ∼ t ( n ) ⇒ X 2 ∼ F ( 1 , n ) \boxed{X\sim t(n)\;\Rightarrow\; X^{2}\sim F(1,n)} X∼t(n)⇒X2∼F(1,n)
E ( X ) E(X) E(X) | D ( X ) D(X) D(X) | |
---|---|---|
X ∼ χ 2 ( n ) X\sim \chi^{2}(n) X∼χ2(n) | n \color{red}n n | 2 n \color{red}2n 2n |
X ∼ t ( n ) X\sim t(n) X∼t(n) | 0 ( n > 1 ) 0\;(n>1) 0(n>1) | n n − 2 ( n > 2 ) \frac{n}{n-2}\;(n>2) n−2n(n>2) |
X ∼ F ( m , n ) X\sim F(m,n) X∼F(m,n) | n n − 2 ( n > 2 ) \frac{n}{n-2}\;(n>2) n−2n(n>2) | 2 n 2 ( m + n − 2 ) m ( n − 2 ) 2 ( n − 4 ) ( n > 4 ) \frac{2n^{2}(m+n-2)}{m(n-2)^{2}(n-4)}\;(n>4) m(n−2)2(n−4)2n2(m+n−2)(n>4) |
(1) 总体( X X X): 与所研究的问题有关的对象(个体)的全体所构成的集合.
——有限总体、无限总体、总体分布
(2) 样本[变量 ( X 1 , ⋯ , X n ) (X_{1},\cdots,X_{n}) (X1,⋯,Xn), 观测值 ( x 1 , ⋯ , x n ) (x_{1},\cdots,x_{n}) (x1,⋯,xn)]: 按一定的规定从总体中抽出的一部分个体.
——简单随机抽样、独立同分布
(3) 统计量: 不含任何未知参数的样本函数.
样本均值: X ‾ = 1 n ∑ i = 1 n X i \overline{X}=\frac{1}{n}\sum_{i=1}^{n}X_{i} X=n1i=1∑nXi样本方差: S 2 = 1 n − 1 ∑ i = 1 n ( X i − X ‾ ) 2 S^{2}=\frac{1}{\color{red} n-1}\sum_{i=1}^{n}(X_{i}-\overline{X})^{2} S2=n−11i=1∑n(Xi−X)2样本标准差: S = S 2 S=\sqrt{S^{2}} S=S2样本 k k k阶(原点)矩: A k = 1 n ∑ i = 1 n X i k ( k ∈ N + ) A_{k}=\frac{1}{n}\sum_{i=1}^{n}X_{i}^{k}\;(k\in \mathbb{N_{+}}) Ak=n1i=1∑nXik(k∈N+)样本 k k k阶中心矩: B k = 1 n ∑ i = 1 n ( X i − X ‾ ) k ( k ∈ N + ) B_{k}=\frac{1}{n}\sum_{i=1}^{n}(X_{i}-\overline{X})^{k}\;(k\in \mathbb{N_{+}}) Bk=n1i=1∑n(Xi−X)k(k∈N+)部分性质: X ‾ 与 S 2 相互独立 \boxed{\overline{X}与S^{2}相互独立} X与S2相互独立 A 1 = X ‾ , B 2 = n − 1 n S 2 \boxed{A_{1}=\overline{X},\; B_{2}=\frac{n-1}{n}S^{2}} A1=X,B2=nn−1S2 总体 X , E ( X ) = μ , D ( X ) = σ 2 , X 1 ⋯ X n 是来自总体 X 的样本 , 则有 E ( X ‾ ) = μ , D ( X ‾ ) = σ 2 n , E ( S 2 ) = σ 2 , D ( S 2 ) = 2 σ 4 n − 1 . \boxed{总体X,\; E(X)=\mu,\; D(X)=\sigma^{2},\;X_{1}\cdots X_{n}是来自总体X的样本,\; 则有E(\overline{X})=\mu,\;D(\overline{X})=\frac{\sigma^{2}}{n},\;E(S^{2})=\sigma^{2},\;D(S^{2})=\frac{2\sigma^{4}}{n-1}.} 总体X,E(X)=μ,D(X)=σ2,X1⋯Xn是来自总体X的样本,则有E(X)=μ,D(X)=nσ2,E(S2)=σ2,D(S2)=n−12σ4.
X 1 , ⋯ , X n X_{1},\cdots,X_{n} X1,⋯,Xn是来自正态总体 X ∼ N ( μ , σ 2 ) X\sim N(\mu, \sigma^{2}) X∼N(μ,σ2)的样本
X ‾ ∼ N ( μ , σ 2 n ) \boxed{\overline X\sim N(\mu,\frac{\sigma^{2}}{n})} X∼N(μ,nσ2) ( n − 1 ) S 2 σ 2 = ∑ i = 1 n ( X i − X ‾ ) 2 σ 2 ∼ χ 2 ( n − 1 ) \boxed{\frac{(n-1)S^{2}}{\sigma^{2}}=\frac{\displaystyle\sum_{i=1}^{n}(X_{i}-\textcolor{blue}{\overline X})^{2}}{\sigma^{2}}\sim \chi^{2}(\textcolor{red}{n-1})} σ2(n−1)S2=σ2i=1∑n(Xi−X)2∼χ2(n−1) ∑ i = 1 n ( X i − μ ) 2 σ 2 ∼ χ 2 ( n ) \frac{\displaystyle\sum_{i=1}^{n}(X_{i}-\textcolor{blue}{\mu})^{2}}{\sigma^{2}}\sim \chi^{2}(\textcolor{red}{n}) σ2i=1∑n(Xi−μ)2∼χ2(n) X ‾ − μ S / n ∼ t ( n − 1 ) \boxed{\frac{\overline{X}-\mu}{\textcolor{blue}{S}/\sqrt{n}}\sim \textcolor{red}{t(n-1)}} S/nX−μ∼t(n−1) X ‾ − μ σ / n ∼ N ( 0 , 1 ) \frac{\overline{X}-\mu}{\textcolor{blue}{\sigma}/\sqrt{n}}\sim \textcolor{red}{N(0,1)} σ/nX−μ∼N(0,1)
X 1 , ⋯ , X n X_{1},\cdots,X_{n} X1,⋯,Xn是来自正态总体 X ∼ N ( μ 1 , σ 1 2 ) X\sim N(\mu_{1}, \sigma_{1}^{2}) X∼N(μ1,σ12)的样本, Y 1 , ⋯ , Y m Y_{1},\cdots,Y_{m} Y1,⋯,Ym是来自正态总体 Y ∼ N ( μ 2 , σ 2 2 ) Y\sim N(\mu_{2}, \sigma_{2}^{2}) Y∼N(μ2,σ22)的样本
X ‾ − Y ‾ − ( μ 1 − μ 2 ) σ 1 2 n + σ 2 2 m ∼ N ( 0 , 1 ) \frac {\overline{X}-\overline{Y}-(\mu_{1}-\mu_{2})} {\sqrt{\frac{\sigma_{1}^{2}}{n}+\frac{\sigma_{2}^{2}}{m}}} \sim N(0,1) nσ12+mσ22X−Y−(μ1−μ2)∼N(0,1) S 2 2 / σ 2 2 S 1 2 / σ 1 2 ∼ F ( m − 1 , n − 1 ) \boxed{\frac{S_{2}^{2}/\sigma_{2}^{2}}{S_{1}^{2}/\sigma_{1}^{2}}\sim F(m-1,n-1)} S12/σ12S22/σ22∼F(m−1,n−1) 当 σ 1 2 = σ 2 2 时 , X ‾ − Y ‾ − ( μ 1 − μ 2 ) ( n − 1 ) S 1 2 + ( m − 1 ) S 2 2 n + m − 2 1 n + 1 m ∼ t ( n + m − 2 ) \boxed{当\sigma_{1}^{2}=\sigma_{2}^{2}时,\;\frac {\overline{X}-\overline{Y}-(\mu_{1}-\mu_{2})} {\sqrt{\frac {(n-1)S_{1}^{2}+(m-1)S_{2}^{2}} {n+m-2} }\sqrt{\frac{1}{n}+\frac{1}{m}}} \sim t(n+m-2)} 当σ12=σ22时,n+m−2(n−1)S12+(m−1)S22n1+m1X−Y−(μ1−μ2)∼t(n+m−2)
总体分布为 f ( x ; θ 1 , ⋯ , θ k ) f(x;\theta_{1},\cdots,\theta_{k}) f(x;θ1,⋯,θk). 解矩估计方程组 α m ( θ 1 , ⋯ , θ k ) = A m \alpha_{m}(\theta_{1},\cdots,\theta_{k})=A_{m} αm(θ1,⋯,θk)=Am, 即: ∫ − ∞ ∞ x m f ( x ; θ 1 , ⋯ , θ k ) d x 或 ∑ i x i m f ( x i ; θ 1 , ⋯ , θ k ) = 1 n ∑ i = 1 n X i m ( m = 1 , ⋯ , k ) \boxed{\int_{-\infty}^{\infty}x^{m}f(x;\theta_{1},\cdots,\theta_{k})\textup{d}x或\sum_{i}x_{i}^{m}f(x_{i};\theta_{1},\cdots,\theta_{k})=\frac{1}{n}\sum_{i=1}^{n}X_{i}^{m}\;(m=1,\cdots,k)} ∫−∞∞xmf(x;θ1,⋯,θk)dx或i∑ximf(xi;θ1,⋯,θk)=n1i=1∑nXim(m=1,⋯,k), 得到 θ j \theta_{j} θj的矩估计值 θ j ^ = θ j ^ ( X 1 , ⋯ , X n ) \hat{\theta_{j}}=\hat{\theta_{j}}(X_{1},\cdots,X_{n}) θj^=θj^(X1,⋯,Xn).
最常见的情况是用样本的 A 1 A_{1} A1 (即 X ‾ \overline{X} X)、 A 2 A_{2} A2 (或 B 2 B_{2} B2)去估计:
(1) 求出总体的矩 μ = E ( X ) = g ( θ ) \mu=E(X)=g(\theta) μ=E(X)=g(θ);
(2) 令 μ = g ( θ ) = X ‾ \mu=g(\theta)=\overline{X} μ=g(θ)=X, 反解出估计量方程 θ ^ = h ( X ‾ ) \hat{\theta}=h(\overline{X}) θ^=h(X);
(3) 将 X ‾ \overline{X} X的值代入 θ ^ = h ( X ‾ ) \hat{\theta}=h(\overline{X}) θ^=h(X)计算出估计值.
有两个未知参数 θ 1 , θ 2 \theta_{1},\theta_{2} θ1,θ2的情况: { X ‾ = μ ; B 2 = σ 2 (对矩估计的修正: S 2 = σ 2 ) . \boxed{\begin{cases} \overline{X}=\mu;\\ B_{2}=\sigma^{2}\;\textup{(对矩估计的修正: }S^{2}=\sigma^{2}). \end{cases}} {X=μ;B2=σ2(对矩估计的修正: S2=σ2).
总体 X X X, P ( X = x ) = p ( x ; θ 1 , ⋯ , θ k ) P(X=x)=p(x;\theta_{1},\cdots,\theta_{k}) P(X=x)=p(x;θ1,⋯,θk)(对离散型总体)或概率密度为 f ( x ; θ 1 , ⋯ , θ k ) f(x;\theta_{1},\cdots,\theta_{k}) f(x;θ1,⋯,θk)(对连续型总体), 其中 θ \theta θ为未知参数. 设 x 1 , ⋯ , x n x_{1},\cdots,x_{n} x1,⋯,xn是一组样本观测值.
(1) 计算似然函数 L ( θ 1 , ⋯ , θ k ) = ∏ i = 1 n p ( x i ; θ 1 , ⋯ , θ k ) 或 ∏ i = 1 n f ( x i ; θ 1 , ⋯ , θ k ) L(\theta_{1},\cdots,\theta_{k})=\displaystyle\prod_{i=1}^{n}p(x_{i};\theta_{1},\cdots,\theta_{k})或\displaystyle\prod_{i=1}^{n}f(x_{i};\theta_{1},\cdots,\theta_{k}) L(θ1,⋯,θk)=i=1∏np(xi;θ1,⋯,θk)或i=1∏nf(xi;θ1,⋯,θk);
(2) 对似然函数(其关于 θ j {\theta_{j}} θj的部分通常较为复杂)取对数得 l ( θ 1 , ⋯ , θ k ) = ln L ( θ 1 , ⋯ , θ k ) = ∑ i = 1 n ln p ( x i ; θ 1 , ⋯ , θ k ) 或 ∑ i = 1 n ln f ( x i ; θ 1 , ⋯ , θ k ) \boxed{l(\theta_{1},\cdots,\theta_{k})=\textup{ln}L(\theta_{1},\cdots,\theta_{k})=\sum_{i=1}^{n}\textup{ln}p(x_{i};\theta_{1},\cdots,\theta_{k})或\sum_{i=1}^{n}\textup{ln}f(x_{i};\theta_{1},\cdots,\theta_{k})} l(θ1,⋯,θk)=lnL(θ1,⋯,θk)=i=1∑nlnp(xi;θ1,⋯,θk)或i=1∑nlnf(xi;θ1,⋯,θk);
(3) 对 θ j \theta_{j} θj求(偏)导, 并令 ∂ ∂ θ j l ( θ 1 , ⋯ , θ k ) = 0 \boxed{\frac{\partial}{\partial\theta_{j}}l(\theta_{1},\cdots,\theta_{k})=0} ∂θj∂l(θ1,⋯,θk)=0, 解出极大似然估计值 θ j ^ \hat{\theta_{j}} θj^;
(4) 验证二阶(偏)导 ∂ 2 ∂ θ j 2 l ( θ 1 , ⋯ , θ k ) ∣ θ j = θ j ^ < 0 \frac{\partial^{2}}{\partial\theta_{j}^{2}}l(\theta_{1},\cdots,\theta_{k})|_{\theta_{j}=\hat{\theta_{j}}}<0 ∂θj2∂2l(θ1,⋯,θk)∣θj=θj^<0, 确保 θ j ^ \hat{\theta_{j}} θj^为区间上的极大值点.
如果发现似然函数 L ( θ 1 , ⋯ , θ k ) L(\theta_{1},\cdots,\theta_{k}) L(θ1,⋯,θk)在区间上关于待估参数 θ j \theta_{j} θj是单调的, 也即方程 ∂ ∂ θ j l ( θ 1 , ⋯ , θ k ) = 0 或 ∂ ∂ θ j L ( θ 1 , ⋯ , θ k ) = 0 \frac{\partial}{\partial\theta_{j}}l(\theta_{1},\cdots,\theta_{k})=0或\frac{\partial}{\partial\theta_{j}}L(\theta_{1},\cdots,\theta_{k})=0 ∂θj∂l(θ1,⋯,θk)=0或∂θj∂L(θ1,⋯,θk)=0无解, 那么就要根据 L ( θ 1 , ⋯ , θ k ) L(\theta_{1},\cdots,\theta_{k}) L(θ1,⋯,θk)关于 θ j \theta_{j} θj的单调性来选取 θ j ^ \hat{\theta_{j}} θj^的值了. 要注意 x x x的取值范围, 以及取值范围与 θ j \theta_{j} θj的关联.
设 θ ^ = θ ^ ( X 1 , ⋯ , X n ) \hat{\theta}=\hat{\theta}(X_{1},\cdots,X_{n}) θ^=θ^(X1,⋯,Xn)是未知参数 θ \theta θ的估计量.
【相合性】
若 ∀ ε > 0 \forall \varepsilon>0 ∀ε>0, 有 lim n → ∞ P ( ∣ θ ^ − θ ∣ ⩾ ε ) = 0 \lim\limits_{n\to \infty}P(|\hat{\theta}-\theta|\geqslant \varepsilon)=0 n→∞limP(∣θ^−θ∣⩾ε)=0 (依概率收敛), 则称 θ ^ \hat{\theta} θ^为 θ \theta θ的相合估计量(一致估计量).
【无偏性】
若 E ( θ ^ ) = θ \boxed{E(\hat{\theta})=\theta} E(θ^)=θ, 则称 θ ^ \hat{\theta} θ^为 θ \theta θ的无偏估计量;
若 lim n → ∞ E ( θ ^ ) = θ \lim\limits_{n\to \infty}E(\hat{\theta})=\theta n→∞limE(θ^)=θ, 则称 θ ^ \hat{\theta} θ^为 θ \theta θ的渐近无偏估计量.
【有效性】
对于待估参数 θ \theta θ的估计量 θ 1 ^ , θ 2 ^ \hat{\theta_{1}},\hat{\theta_{2}} θ1^,θ2^, 若 D ( θ 1 ^ ) ⩽ D ( θ 2 ^ ) \boxed{D(\hat{\theta_{1}})\leqslant D(\hat{\theta_{2}})} D(θ1^)⩽D(θ2^), 则称 θ 1 ^ \hat{\theta_{1}} θ1^比 θ 2 ^ \hat{\theta_{2}} θ2^更加有效.
P ( θ ‾ < θ < θ ‾ ) = 1 − α ( 0 < α < 1 ) P(\underline{\theta}<\theta<\overline{\theta})=1-\alpha\; (0<\alpha<1) P(θ<θ<θ)=1−α(0<α<1)
随机区间 ( θ ‾ , θ ‾ ) (\underline{\theta},\overline{\theta}) (θ,θ)是 θ \theta θ的置信度为 1 − α 1-\alpha 1−α的置信区间, 分别称 θ ‾ \underline{\theta} θ和 θ ‾ \overline{\theta} θ为置信度为 1 − α 1-\alpha 1−α的置信上限和置信下限.
u α u_{\alpha} uα是 X X X或其分布的上 α \alpha α分位点: P ( X > u α ) = α P(X>u_{\alpha})=\alpha P(X>uα)=α
记号约定: N ( 0 , 1 ) − z α N(0,1)-\color{red}z_{\alpha} N(0,1)−zα; χ 2 ( n ) − χ α 2 ( n ) \chi^{2}(n)-\color{red}\chi_{\alpha}^{2}(n) χ2(n)−χα2(n); t ( n ) − t α ( n ) t(n)-\color{red}t_{\alpha}(n) t(n)−tα(n); F ( n , m ) − F α ( n , m ) F(n,m)-\color{red}F_{\alpha}(n,m) F(n,m)−Fα(n,m)
性质: z 1 − α = − z α z_{1-\alpha}=-z_{\alpha} z1−α=−zα; t 1 − α ( n ) = − t α ( n ) t_{1-\alpha}(n)=-t_{\alpha}(n) t1−α(n)=−tα(n); F α ( n , m ) = 1 F 1 − α ( m , n ) F_{\alpha}(n,m)=\displaystyle{\frac{1}{F_{1-\alpha}(m,n)}} Fα(n,m)=F1−α(m,n)1
H 0 H_{0} H0 | H 1 H_{1} H1 | 检验统计量 | 拒绝域 W W W |
---|---|---|---|
μ = μ 0 \mu=\mu_{0} μ=μ0 | μ ≠ μ 0 \mu\ne\mu_{0} μ=μ0 | x ‾ − μ 0 σ / n \frac{\overline{x}-\mu_{0}}{\sigma/\sqrt{n}} σ/nx−μ0 | { ∣ x ‾ − μ 0 σ / n ∣ > z α 2 } \left\{\left\vert\frac{\overline{x}-\mu_{0}}{\sigma/\sqrt{n}}\right\vert>z_{\frac{\alpha}{2}}\right\} { σ/nx−μ0 >z2α} |
μ = μ 0 / μ ⩽ μ 0 \mu=\mu_{0}/\mu\leqslant\mu_{0} μ=μ0/μ⩽μ0 | μ > μ 0 \mu>\mu_{0} μ>μ0 | x ‾ − μ 0 σ / n \frac{\overline{x}-\mu_{0}}{\sigma/\sqrt{n}} σ/nx−μ0 | { ∣ x ‾ − μ 0 σ / n ∣ > z α } \left\{\left\vert\frac{\overline{x}-\mu_{0}}{\sigma/\sqrt{n}}\right\vert>z_{\alpha}\right\} { σ/nx−μ0 >zα} |
μ = μ 0 / μ ⩾ μ 0 \mu=\mu_{0}/\mu\geqslant\mu_{0} μ=μ0/μ⩾μ0 | μ < μ 0 \mu<\mu_{0} μ<μ0 | x ‾ − μ 0 σ / n \frac{\overline{x}-\mu_{0}}{\sigma/\sqrt{n}} σ/nx−μ0 | { ∣ x ‾ − μ 0 σ / n ∣ < − z α } \left\{\left\vert\frac{\overline{x}-\mu_{0}}{\sigma/\sqrt{n}}\right\vert<-z_{\alpha}\right\} { σ/nx−μ0 <−zα} |
H 0 H_{0} H0 | H 1 H_{1} H1 | 检验统计量 | 拒绝域 W W W |
---|---|---|---|
μ = μ 0 \mu=\mu_{0} μ=μ0 | μ ≠ μ 0 \mu\ne\mu_{0} μ=μ0 | x ‾ − μ 0 S / n \frac{\overline{x}-\mu_{0}}{S/\sqrt{n}} S/nx−μ0 | { ∣ x ‾ − μ 0 S / n ∣ > t α 2 ( n − 1 ) } \left\{\left\vert\frac{\overline{x}-\mu_{0}}{S/\sqrt{n}}\right\vert>t_{\frac{\alpha}{2}}(n-1)\right\} { S/nx−μ0 >t2α(n−1)} |
μ = μ 0 / μ ⩽ μ 0 \mu=\mu_{0}/\mu\leqslant\mu_{0} μ=μ0/μ⩽μ0 | μ > μ 0 \mu>\mu_{0} μ>μ0 | x ‾ − μ 0 S / n \frac{\overline{x}-\mu_{0}}{S/\sqrt{n}} S/nx−μ0 | { ∣ x ‾ − μ 0 S / n ∣ > t α ( n − 1 ) } \left\{\left\vert\frac{\overline{x}-\mu_{0}}{S/\sqrt{n}}\right\vert>t_{\alpha}(n-1)\right\} { S/nx−μ0 >tα(n−1)} |
μ = μ 0 / μ ⩾ μ 0 \mu=\mu_{0}/\mu\geqslant\mu_{0} μ=μ0/μ⩾μ0 | μ < μ 0 \mu<\mu_{0} μ<μ0 | x ‾ − μ 0 S / n \frac{\overline{x}-\mu_{0}}{S/\sqrt{n}} S/nx−μ0 | { ∣ x ‾ − μ 0 S / n ∣ < − t α ( n − 1 ) } \left\{\left\vert\frac{\overline{x}-\mu_{0}}{S/\sqrt{n}}\right\vert<-t_{\alpha}(n-1)\right\} { S/nx−μ0 <−tα(n−1)} |
成对数据 x 1 , ⋯ , x n x_{1},\cdots,x_{n} x1,⋯,xn和 y 1 , ⋯ , y n y_{1},\cdots,y_{n} y1,⋯,yn
配对差值 d i = y i − x i ( i = 1 , ⋯ , n ) d_{i}=y_{i}-x_{i}\;(i=1,\cdots,n) di=yi−xi(i=1,⋯,n)
可以将 d 1 , ⋯ , d n d_{1},\cdots,d_{n} d1,⋯,dn视为来自正态总体 N ( μ , σ 2 ) N(\mu,\sigma^{2}) N(μ,σ2)的样本
问题变成对单个正态总体在方差未知时, 对均值的 t t t检验