宋浩概率论与数理统计-第七章-笔记

概率论与数理统计

  • 第七章
    • 7.1 点估计
      • 7.1.1 矩估计法
      • 7.1.2 极大似然估计法
    • 7.2 点估计的优良性准则
    • 7.3
      • 7.3.1 置信区间
      • 7.3.2 一个正态总体的均值和方差的区间估计
        • 期望的区间估计
        • 方差的区间估计
        • 总结

第七章

7.1 点估计

估计:已知分布,提取样本,构造函数估计参数的值( θ ^ = θ ^ ( X 1 , ⋯   , X n ) \hat{\theta}=\hat{\theta}(X_1,\cdots,X_n) θ^=θ^(X1,,Xn)
参数空间:参数的取值范围

7.1.1 矩估计法

总体思想:用样本的矩代替总体的矩

总体的矩 ⟵ \longleftarrow 样本的矩
一阶 E X ⟵ EX\longleftarrow EX 一阶 X ‾ = 1 n ∑ X i \overline{X}=\frac{1}{n}\sum X_i X=n1Xi
二阶 E X 2 ⟵ EX^2\longleftarrow EX2 二阶 A 2 = 1 n ∑ X i 2 A_2=\frac{1}{n}\sum X_i^2 A2=n1Xi2

例题
【例1】 X ∼ N ( μ , σ 2 ) X\sim N(\mu,\sigma^2) XN(μ,σ2) ( X 1 , ⋯   , X n ) (X_1,\cdots,X_n) (X1,,Xn)是样本,用矩估计法估计 μ , σ 2 \mu,\sigma^2 μ,σ2
解:
总体一阶矩: E X = μ EX=\mu EX=μ
样本一阶矩: X ‾ = 1 n ∑ i = 1 n X i \displaystyle\overline{X}=\frac{1}{n}\sum\limits_{i=1}^nX_i X=n1i=1nXi
μ ^ = X ‾ \hat{\mu}=\overline{X} μ^=X

总体二阶矩: E X 2 = D X + ( E X ) 2 = σ 2 + μ 2 EX^2=DX+(EX)^2=\sigma^2+\mu^2 EX2=DX+(EX)2=σ2+μ2x
样本二阶矩: A 2 = 1 n ∑ i = 1 n X i 2 \displaystyle A_2=\frac{1}{n}\sum\limits_{i=1}^{n}X_i^2 A2=n1i=1nXi2
σ ^ 2 = A 2 − μ ^ 2 = 1 n ∑ i = 1 n X i 2 − X ‾ 2 = 1 n ∑ i = 1 n ( X i − X ‾ ) 2 = B 2 \displaystyle\hat{\sigma}^2=A_2-\hat{\mu}^2=\frac{1}{n}\sum\limits_{i=1}^{n}X_i^2-\overline{X}^2=\frac{1}{n}\sum\limits_{i=1}^n(X_i-\overline{X})^2=B_2 σ^2=A2μ^2=n1i=1nXi2X2=n1i=1n(XiX)2=B2

【例2】 X ∼ P ( λ ) X\sim P(\lambda) XP(λ) ( X 1 , ⋯   , X n ) (X_1,\cdots,X_n) (X1,,Xn)是样本,用矩估计法估计
解:
E X = λ EX=\lambda EX=λ
λ ^ = X ‾ \hat{\lambda}=\overline{X} λ^=X

D X = λ DX=\lambda DX=λ
λ ^ = B 2 \hat{\lambda}=B_2 λ^=B2

【例3】 X X X服从 [ θ 1 , θ 2 ] [\theta_1,\theta_2] [θ1,θ2]上的均匀分布, ( X 1 , ⋯   , X n ) (X_1,\cdots,X_n) (X1,,Xn)是样本,用矩估计法估计
解:
E X = 1 2 ( θ 1 , θ 2 ) = X ‾ \displaystyle EX=\frac{1}{2}(\theta_1,\theta_2)=\overline{X} EX=21(θ1,θ2)=X
E X 2 = D X + ( E X ) 2 = ( θ 1 − θ 2 ) 2 12 + ( θ 1 + θ 2 ) 2 4 = A 2 \displaystyle EX^2=DX+(EX)^2=\frac{(\theta_1-\theta_2)^2}{12}+\frac{(\theta_1+\theta_2)^2}{4}=A_2 EX2=DX+(EX)2=12(θ1θ2)2+4(θ1+θ2)2=A2

解上述方程得:
θ 1 ^ = X ‾ − 3 B 2 , θ 2 ^ = X ‾ + 3 B 2 \hat{\theta_1}=\overline{X}-\sqrt{3B_2},\hat{\theta_2}=\overline{X}+\sqrt{3B_2} θ1^=X3B2 ,θ2^=X+3B2

【例4】 X X X满足:
X 1 2 3 P θ 2 2 θ ( 1 − θ ) ( 1 − θ ) 2 \begin{array}{c|c} X & 1 & 2 & 3\\ \hline P & \theta^2 & 2\theta(1-\theta) & (1-\theta)^2 \end{array} XP1θ222θ(1θ)3(1θ)2
取样本 { 1 , 2 , 1 } \{1,2,1\} {1,2,1},求矩估计
解:
E X = 1 × θ + 2 × 2 θ ( 1 − θ ) + 3 × ( 1 − θ ) 2 EX=1\times\theta+2\times2\theta(1-\theta)+3\times(1-\theta)^2 EX=1×θ+2×2θ(1θ)+3×(1θ)2
X ‾ = 1 + 2 + 1 3 = 4 3 \displaystyle \overline{X}=\frac{1+2+1}{3}=\frac{4}{3} X=31+2+1=34
θ ^ = 5 6 \displaystyle\hat{\theta}=\frac{5}{6} θ^=65

矩估计:简单易行,在使用时并不需要事先知道总体的分布。但是矩估计没有充分利用总体所提供的信息,因此矩估计不一定时理想的估计,另外,矩估计也有不唯一性。矩估计应用的前提时总体的矩存在。

7.1.2 极大似然估计法

举例:
100 100 100个球,黑色 99 99 99个,白色 1 1 1 ⟹ \Longrightarrow 摸到黑球的概率为 0.99 0.99 0.99,摸到白球的概率为 0.01 0.01 0.01
100 100 100个球,黑球个数 θ = 99 或 1 \theta=99或1 θ=991,白球个数 = 99 或 1 =99或1 =991,任意摸出一个是黑球 ⟹ θ \Longrightarrow\theta θ很可能为 99 99 99

P P P大的事件比 P P P小的事件更容易发生

将使得 A A A发生的概率最大的参数值作为估计值

【解题模板】:

  1. 总体的概率函数(离散型)或密度函数(连续型)
  2. 写似然函数 L ( λ ) L(\lambda) L(λ)
  3. 两边取 ln ⁡ \ln ln ln ⁡ L ( λ ) \ln L(\lambda) lnL(λ)
  4. λ \lambda λ求(偏)导,令偏导数=0

例题
【例2】总体 X ∼ P ( λ ) X\sim P(\lambda) XP(λ) ( X 1 , X 2 , ⋯   , X n ) (X_1,X_2,\cdots,X_n) (X1,X2,,Xn)为样本,求 λ \lambda λ的极大似然估计
解:
总体的概率函数为: P ( X = k ) = λ k k ! e − λ ( k = 0 , 1 , 2 , ⋯   ) P(X=k)=\displaystyle\frac{\lambda^k}{k!}e^{-\lambda}(k=0,1,2,\cdots) P(X=k)=k!λkeλ(k=0,1,2,)

λ \lambda λ的似然函数为: L ( λ ) = ∏ i = 1 n λ x i x i ! e − λ = λ x 1 + x 2 + ⋯ + x n ∏ i = 1 n x i ! e − n λ L(\lambda)=\displaystyle\prod\limits_{i=1}^n\frac{\lambda^{x_i}}{x_i!}e^{-\lambda}=\frac{\lambda^{x_1+x_2+\cdots+x_n}}{\prod\limits_{i=1}^nx_i!}e^{-n\lambda} L(λ)=i=1nxi!λxieλ=i=1nxi!λx1+x2++xnenλ
x i x_i xi:观测值, λ \lambda λ:未知数)
两边取 ln ⁡ \ln ln ln ⁡ L ( λ ) = − ln ⁡ ∏ i = 1 n x i ! + ( x 1 + ⋯ + x n ) ln ⁡ λ − n λ \displaystyle\ln L(\lambda)=-\ln\prod\limits_{i=1}^nx_i!+(x_1+\cdots+x_n)\ln\lambda-n\lambda lnL(λ)=lni=1nxi!+(x1++xn)lnλnλ

两边对 λ \lambda λ求导: d ln ⁡ L ( λ ) d λ = x 1 + ⋯ + x n λ − n \displaystyle\frac{d\ln L(\lambda)}{d\lambda}=\frac{x_1+\cdots+x_n}{\lambda}-n dλdlnL(λ)=λx1++xnn

d ln ⁡ L ( λ ) d λ = 0 \displaystyle\frac{d\ln L(\lambda)}{d\lambda}=0 dλdlnL(λ)=0

λ = x 1 + ⋯ + x n n = x ‾ \lambda=\displaystyle\frac{x_1+\cdots+x_n}{n}=\overline{x} λ=nx1++xn=x

【例3】总体是参数为 λ \lambda λ的指数分布, ( X 1 , X 2 , ⋯   , X n ) (X_1,X_2,\cdots,X_n) (X1,X2,,Xn)为样本,求 λ \lambda λ的极大似然估计
解:
总体的密度函数为: f ( x ; λ ) = { λ e − λ x x > 0 0 x ≤ 0 f(x;\lambda)=\begin{cases} \lambda e^{-\lambda x} & x>0\\ 0 & x\leq0 \end{cases} f(x;λ)={λeλx0x>0x0

λ \lambda λ的似然函数为: L ( λ ) = ∏ i = 1 n λ e − λ x i = λ n e − λ ( x 1 + ⋯ + x n ) L(\lambda)=\displaystyle\prod\limits_{i=1}^{n}\lambda e^{-\lambda x_i}=\lambda^{n}e^{-\lambda(x_1+\cdots+x_n)} L(λ)=i=1nλeλxi=λneλ(x1++xn)
两边取 ln ⁡ \ln ln ln ⁡ L ( λ ) = n ln ⁡ λ − λ ( x 1 + ⋯ + x n ) \displaystyle\ln L(\lambda)=n\ln \lambda-\lambda(x_1+\cdots+x_n) lnL(λ)=nlnλλ(x1++xn)

两边对 λ \lambda λ求导: d ln ⁡ L ( λ ) d λ = n λ − ( x 1 + ⋯ + x n ) \displaystyle\frac{d\ln L(\lambda)}{d\lambda}=\frac{n}{\lambda}-(x_1+\cdots+x_n) dλdlnL(λ)=λn(x1++xn)

d ln ⁡ L ( λ ) d λ = 0 \displaystyle\frac{d\ln L(\lambda)}{d\lambda}=0 dλdlnL(λ)=0

λ = n x 1 + ⋯ + x n = 1 x ‾ \lambda=\displaystyle\frac{n}{x_1+\cdots+x_n}=\frac{1}{\overline{x}} λ=x1++xnn=x1

【例4】总体 X ∼ N ( μ , σ 2 ) X\sim N(\mu,\sigma^2) XN(μ,σ2) ( X 1 , X 2 , ⋯   , X n ) (X_1,X_2,\cdots,X_n) (X1,X2,,Xn)为样本,求 μ , σ 2 \mu,\sigma^2 μ,σ2的极大似然估计
解:
总体的密度函数为: f ( x ; μ , σ 2 ) = 1 2 π σ e − ( x − μ ) 2 2 σ 2 f(x;\mu,\sigma^2)=\displaystyle\frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{(x-\mu)^2}{2\sigma^2}} f(x;μ,σ2)=2π σ1e2σ2(xμ)2

μ , σ 2 \mu,\sigma^2 μ,σ2的似然函数为: L ( μ , σ 2 ) = ∏ i = 1 n 1 2 π σ e − ( x i − μ ) 2 2 σ 2 = ( 1 2 π ) n ( 1 σ ) n e − ( x 1 − μ ) 2 + ( x 2 − μ ) 2 + ⋯ + ( x n − μ ) 2 2 σ 2 L(\mu,\sigma^2)=\displaystyle\prod\limits_{i=1}^{n}\frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{(x_i-\mu)^2}{2\sigma^2}}=(\frac{1}{\sqrt{2\pi}})^n(\frac{1}{\sigma})^ne^{-\frac{(x_1-\mu)^2+(x_2-\mu)^2+\cdots+(x_n-\mu)^2}{2\sigma^2}} L(μ,σ2)=i=1n2π σ1e2σ2(xiμ)2=(2π 1)n(σ1)ne2σ2(x1μ)2+(x2μ)2++(xnμ)2
两边取 ln ⁡ \ln ln ln ⁡ L ( μ , σ 2 ) = n ln ⁡ 1 2 π − n 2 ln ⁡ σ 2 − ( x 1 − μ ) 2 + ( x 2 − μ ) 2 + ⋯ + ( x n − μ ) 2 2 σ 2 \displaystyle\ln L(\mu,\sigma^2)=n\ln\frac{1}{\sqrt{2\pi}}-\frac{n}{2}\ln\sigma^2-\frac{(x_1-\mu)^2+(x_2-\mu)^2+\cdots+(x_n-\mu)^2}{2\sigma^2} lnL(μ,σ2)=nln2π 12nlnσ22σ2(x1μ)2+(x2μ)2++(xnμ)2
( 1 σ ) n = ( σ 2 ) − n 2 (\frac{1}{\sigma})^n=(\sigma^2)^{-\frac{n}{2}} (σ1)n=(σ2)2n

两边对 μ \mu μ求偏导:
∂ ln ⁡ L ( μ , σ 2 ) ∂ μ = − − ( 2 ( x i − μ ) + ⋯ + 2 ( x i − μ ) ) 2 σ 2 = x 1 + ⋯ + x n − n μ σ 2 \displaystyle\frac{\partial\ln L(\mu,\sigma^2)}{\partial\mu}=-\frac{-(2(x_i-\mu)+\cdots+2(x_i-\mu))}{2\sigma^2}=\frac{x_1+\cdots+x_n-n\mu}{\sigma^2} μlnL(μ,σ2)=2σ2(2(xiμ)++2(xiμ))=σ2x1++xnnμ

∂ ln ⁡ L ( μ , σ 2 ) ∂ μ = 0 \displaystyle\frac{\partial\ln L(\mu,\sigma^2)}{\partial\mu}=0 μlnL(μ,σ2)=0

μ = x 1 + ⋯ + x n n = x ‾ \mu=\displaystyle\frac{x_1+\cdots+x_n}{n}=\overline{x} μ=nx1++xn=x

两边对 σ 2 \sigma^2 σ2求偏导:
∂ ln ⁡ L ( μ , σ 2 ) ∂ σ 2 = − n 2 1 σ 2 + ( x 1 − μ ) 2 + ( x 2 − μ ) 2 + ⋯ + ( x n − μ ) 2 2 σ 4 \displaystyle\frac{\partial\ln L(\mu,\sigma^2)}{\partial\sigma^2}=-\frac{n}{2}\frac{1}{\sigma^2}+\frac{(x_1-\mu)^2+(x_2-\mu)^2+\cdots+(x_n-\mu)^2}{2\sigma^4} σ2lnL(μ,σ2)=2nσ21+2σ4(x1μ)2+(x2μ)2++(xnμ)2

∂ ln ⁡ L ( μ , σ 2 ) ∂ σ 2 = 0 \displaystyle\frac{\partial\ln L(\mu,\sigma^2)}{\partial\sigma^2}=0 σ2lnL(μ,σ2)=0

σ 2 = ( x 1 − μ ) 2 + ( x 2 − μ ) 2 + ⋯ + ( x n − μ ) 2 n = B 2 \displaystyle\sigma^2=\frac{(x_1-\mu)^2+(x_2-\mu)^2+\cdots+(x_n-\mu)^2}{n}=B_2 σ2=n(x1μ)2+(x2μ)2++(xnμ)2=B2

【例4】总体 X ∼ [ θ 1 , θ 2 ] X\sim[\theta_1,\theta_2] X[θ1,θ2] ( X 1 , ⋯   , X 2 ) (X_1,\cdots,X_2) (X1,,X2)为样本,求 θ 1 , θ 2 \theta_1,\theta_2 θ1,θ2的极大似然估计
解:
密度函数: f ( x ) = { 1 θ 2 − θ 1 x ∈ [ θ 1 , θ 2 ] 0 e l s e f(x)=\begin{cases} \displaystyle\frac{1}{\theta_2-\theta_1} & x\in[\theta_1,\theta_2]\\ 0 & else \end{cases} f(x)=θ2θ110x[θ1,θ2]else

似然函数: L ( θ 1 , θ 2 ) = ∏ i = 1 n 1 θ 2 − θ 1 = 1 ( θ 2 − θ 1 ) n \displaystyle L(\theta_1,\theta_2)=\prod\limits_{i=1}^{n}\frac{1}{\theta_2-\theta_1}=\frac{1}{(\theta_2-\theta_1)^n} L(θ1,θ2)=i=1nθ2θ11=(θ2θ1)n1

L ( θ 1 , θ 2 ) L(\theta_1,\theta_2) L(θ1,θ2)取最大 ⟹ \Longrightarrow θ 1 , θ 2 \theta_1,\theta_2 θ1,θ2尽量接近

θ 1 = min ⁡ { x 1 , x 2 , ⋯   , x n } \theta_1=\min\{x_1,x_2,\cdots,x_n\} θ1=min{x1,x2,,xn}
θ 2 = max ⁡ { x 1 , x 2 , ⋯   , x n } \theta_2=\max\{x_1,x_2,\cdots,x_n\} θ2=max{x1,x2,,xn}

7.2 点估计的优良性准则

【无偏性】: E ( θ ^ ) = θ E(\hat{\theta})=\theta E(θ^)=θ

  1. (任何分布)总体 X X X E ( X ) = μ , D ( X ) = σ 2 E(X)=\mu,D(X)=\sigma^2 E(X)=μ,D(X)=σ2,样本 ( X 1 , ⋯   , X n ) (X_1,\cdots,X_n) (X1,,Xn)
  • X ‾ \overline{X} X μ \mu μ的无偏估计
  • 样本方差 s 2 s^2 s2 σ 2 \sigma^2 σ2的无偏估计
  1. θ ^ \hat{\theta} θ^ θ \theta θ的无偏估计,则 g ( θ ^ ) g(\hat{\theta}) g(θ^)不一定是 g ( θ ) g(\theta) g(θ)的无偏估计
    s 2 s^2 s2 σ 2 \sigma^2 σ2的无偏估计,但 s s s不是 σ \sigma σ的无偏估计

  2. μ = E ( X ) \mu=E(X) μ=E(X),样本 ( X 1 , ⋯   , X n ) (X_1,\cdots,X_n) (X1,,Xn),取 μ ^ = C 1 X 1 + ⋯ + C n X n \hat{\mu}=C_1X_1+\cdots+C_nX_n μ^=C1X1++CnXn,其中 C 1 + ⋯ + C n = 1 C_1+\cdots+C_n=1 C1++Cn=1,则 μ ^ \hat{\mu} μ^ μ \mu μ的无偏估计

【有效性】: D ( θ ^ 1 ) ≤ D ( θ ^ 2 ) D(\hat{\theta}_1)\leq D(\hat{\theta}_2) D(θ^1)D(θ^2)

  1. 总体 X X X E ( X ) = μ , D ( X ) = σ 2 E(X)=\mu,D(X)=\sigma^2 E(X)=μ,D(X)=σ2 μ = { X 1 E ( X 1 ) = μ D ( X 1 ) = σ 2 X ‾ E ( X ‾ ) = μ D ( X ‾ ) = σ 2 / n 更 有 效 \mu=\begin{cases} X_1 & E(X_1)=\mu & D(X_1)=\sigma^2 &\\ \overline{X} & E(\overline{X})=\mu & D(\overline{X})=\sigma^2/n & 更有效\\ \end{cases} μ={X1XE(X1)=μE(X)=μD(X1)=σ2D(X)=σ2/n
  2. 总体 X X X E ( X ) = μ , D ( X ) = σ 2 E(X)=\mu,D(X)=\sigma^2 E(X)=μ,D(X)=σ2 μ = { a 1 X 1 + ⋯ + a n X n D ( a 1 X 1 + ⋯ + a n X n ) = σ 2 ( a 1 2 + ⋯ + a n 2 ) X ‾ D ( X ‾ ) = σ 2 / n 更 有 效 \mu=\begin{cases} a_1X_1+\cdots+a_nX_n & D(a_1X_1+\cdots+a_nX_n)=\sigma^2(a_1^2+\cdots+a_n^2) &\\ \overline{X} & D(\overline{X})=\sigma^2/n & 更有效\\ \end{cases} μ={a1X1++anXnXD(a1X1++anXn)=σ2(a12++an2)D(X)=σ2/n

【相合性(一致性)】: lim ⁡ n → + ∞ P ( ∣ θ ^ − θ ∣ < ϵ ) = 1 \lim\limits_{n\to+\infin}P(|\hat{\theta}-\theta|<\epsilon)=1 n+limP(θ^θ<ϵ)=1

7.3

7.3.1 置信区间

区间估计:
P ( θ ^ 1 ≤ θ ≤ θ ^ 2 ) = 1 − α P(\hat{\theta}_1\leq\theta\leq\hat{\theta}_2)=1-\alpha P(θ^1θθ^2)=1α
1 − α 1-\alpha 1α:置信度, [ θ ^ 1 , θ ^ 2 ] [\hat{\theta}_1,\hat{\theta}_2] [θ^1,θ^2]:区间)

做题:已知置信度,求 θ ^ 1 , θ ^ 2 \hat{\theta}_1,\hat{\theta}_2 θ^1,θ^2 θ \theta θ为未知参数

置信度: [ θ ^ 1 , θ ^ 2 ] [\hat{\theta}_1,\hat{\theta}_2] [θ^1,θ^2]能套住 θ \theta θ的概率( θ \theta θ是未知的、确定的)

枢轴变量:

  1. 变量 I = I ( T , θ ) I=I(T,\theta) I=I(T,θ) θ \theta θ是未知的参数, T T T已知, I I I的分布 F F F已知且与 θ \theta θ无关, I I I称为枢轴变量
  2. 给定 1 − α 1-\alpha 1α,确定 F F F的上 α 2 \displaystyle\frac{\alpha}{2} 2α分位数 v α 2 \displaystyle v_{\frac{\alpha}{2}} v2α,上 1 − α 2 \displaystyle1-\frac{\alpha}{2} 12α分位数 v 1 − α 2 \displaystyle v_{1-\frac{\alpha}{2}} v12α P ( v 1 − α 2 ≤ I ( T , θ ) ≤ v α 2 ) = 1 − α P(\displaystyle v_{1-\frac{\alpha}{2}}\leq I(T,\theta)\leq v_{\frac{\alpha}{2}})=1-\alpha P(v12αI(T,θ)v2α)=1α

7.3.2 一个正态总体的均值和方差的区间估计

期望的区间估计

  • σ 2 \sigma^2 σ2已知,估计 μ \mu μ(即 μ \mu μ为未知参数)
    构造枢轴变量 U = n ( X ‾ − μ ) σ ∼ N ( 0 , 1 ) U=\displaystyle\frac{\sqrt{n}(\overline{X}-\mu)}{\sigma}\sim N(0,1) U=σn (Xμ)N(0,1)
    σ , n , X ‾ \sigma,n,\overline{X} σ,n,X已知, μ \mu μ未知)
    给定 1 − α 1-\alpha 1α,令 P ( U > u α 2 ) = α 2 , Φ 0 ( u α 2 ) = 1 − α 2 \displaystyle P(U>u_{\frac{\alpha}{2}})=\frac{\alpha}{2},\Phi_0(u_{\frac{\alpha}{2}})=1-\frac{\alpha}{2} P(U>u2α)=2α,Φ0(u2α)=12α
    P ( − u α 2 ≤ n ( X ‾ − μ ) σ ≤ u α 2 ) = 1 − α \displaystyle P(-u_{\frac{\alpha}{2}}\leq\frac{\sqrt{n}(\overline{X}-\mu)}{\sigma}\leq u_{\frac{\alpha}{2}})=1-\alpha P(u2ασn (Xμ)u2α)=1α
    P ( X ‾ − σ u α 2 n ≤ u ≤ X ‾ + σ u α 2 n ) = 1 − α \displaystyle P(\overline{X}-\frac{\sigma u_{\frac{\alpha}{2}}}{\sqrt{n}}\leq u\leq \overline{X}+\frac{\sigma u_{\frac{\alpha}{2}}}{\sqrt{n}})=1-\alpha P(Xn σu2αuX+n σu2α)=1α

例题
【例1】5个灯泡,样本数据:1650,1700,1680,1820,1800, X ∼ N ( μ , 9 ) X\sim N(\mu,9) XN(μ,9) α = 0.05 \alpha=0.05 α=0.05

解:
n = 5 n=5 n=5
X ‾ = 1650 + 1700 + 1680 + 1820 + 1800 5 = 1730 \displaystyle\overline{X}=\frac{1650+1700+1680+1820+1800}{5}=1730 X=51650+1700+1680+1820+1800=1730
σ 2 = 9 ⇒ σ = 3 \sigma^2=9\Rightarrow \sigma=3 σ2=9σ=3

查表 u 0.975 = 1.96 u_{0.975}=1.96 u0.975=1.96

− 1.96 ≤ n ( X ‾ − μ ) σ ≤ 1.96 \displaystyle-1.96\leq\frac{\sqrt{n}(\overline{X}-\mu)}{\sigma}\leq1.96 1.96σn (Xμ)1.96
− 1.96 ≤ 5 ( 1730 − μ ) 3 ≤ 1.96 \displaystyle-1.96\leq\frac{\sqrt{5}(1730-\mu)}{3}\leq1.96 1.9635 (1730μ)1.96
1727.37 ≤ μ ≤ 1732.63 1727.37\leq\mu\leq1732.63 1727.37μ1732.63

  • σ 2 \sigma^2 σ2未知(不可用),估计 μ \mu μ
    构造枢轴变量 T = n ( X ‾ − μ ) s ∼ t ( n − 1 ) \displaystyle T=\frac{\sqrt{n}(\overline{X}-\mu)}{s}\sim t(n-1) T=sn (Xμ)t(n1)
    s s s为样本标准差)

给定 1 − α 1-\alpha 1α,上 α 2 \displaystyle\frac{\alpha}{2} 2α分位数 t α 2 ( n − 1 ) \displaystyle t_{\frac{\alpha}{2}}(n-1) t2α(n1)
P ( − t α 2 ( n − 1 ) ≤ n ( X ‾ − μ ) s ≤ t α 2 ( n − 1 ) ) = 1 − α P(\displaystyle -t_{\frac{\alpha}{2}}(n-1)\leq \displaystyle\frac{\sqrt{n}(\overline{X}-\mu)}{s}\leq t_{\frac{\alpha}{2}}(n-1))=1-\alpha P(t2α(n1)sn (Xμ)t2α(n1))=1α

X ‾ − s n t α 2 ( n − 1 ) ≤ μ ≤ X ‾ + s n t α 2 ( n − 1 ) \displaystyle\overline{X}-\frac{s}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1)\leq\mu\leq\overline{X}+\frac{s}{\sqrt{n}}t_{\frac{\alpha}{2}}(n-1) Xn st2α(n1)μX+n st2α(n1)

方差的区间估计

  1. μ \mu μ已知,对 σ 2 \sigma^2 σ2区间估计
    χ 2 = 1 σ 2 ∑ i = 1 n ( X i − μ ) 2 \displaystyle\chi^2=\frac{1}{\sigma^2}\sum\limits_{i=1}^{n}(X_i-\mu)^2 χ2=σ21i=1n(Xiμ)2
    给定 1 − α 1-\alpha 1α χ 1 − α 2 2 ( n ) , χ α 2 2 ( n ) \displaystyle\chi_{1-\frac{\alpha}{2}}^{2}(n),\chi_{\frac{\alpha}{2}}^{2}(n) χ12α2(n),χ2α2(n)
    χ 1 − α 2 2 ( n ) ≤ 1 σ 2 ∑ i = 1 n ( X i − μ ) 2 ≤ χ α 2 2 ( n ) \displaystyle \chi_{1-\frac{\alpha}{2}}^{2}(n)\leq\frac{1}{\sigma^2}\sum\limits_{i=1}^{n}(X_i-\mu)^2\leq\chi_{\frac{\alpha}{2}}^{2}(n) χ12α2(n)σ21i=1n(Xiμ)2χ2α2(n)
    ∑ i = 1 n ( X i − μ ) 2 χ α 2 2 ( n ) ≤ σ 2 ≤ ∑ i = 1 n ( X i − μ ) 2 χ 1 − α 2 2 ( n ) \displaystyle\frac{\sum\limits_{i=1}^{n}(X_i-\mu)^2}{\chi_{\frac{\alpha}{2}}^{2}(n)}\leq\sigma^2\leq\frac{\sum\limits_{i=1}^{n}(X_i-\mu)^2}{\chi_{1-\frac{\alpha}{2}}^{2}(n)} χ2α2(n)i=1n(Xiμ)2σ2χ12α2(n)i=1n(Xiμ)2

  2. μ \mu μ未知,估计 σ 2 \sigma^2 σ2
    χ 2 = ( n − 1 ) s 2 σ 2 ∼ χ 2 ( n − 1 ) \displaystyle\chi^2=\frac{(n-1)s^2}{\sigma^2}\sim\chi^2(n-1) χ2=σ2(n1)s2χ2(n1)
    给定 1 − α 1-\alpha 1α χ 1 − α 2 2 ( n − 1 ) , χ α 2 2 ( n − 1 ) \displaystyle\chi_{1-\frac{\alpha}{2}}^{2}(n-1),\chi_{\frac{\alpha}{2}}^{2}(n-1) χ12α2(n1),χ2α2(n1)
    置信区间: [ ( n − 1 ) s 2 χ α 2 2 ( n − 1 ) , ( n − 1 ) s 2 χ 1 − α 2 2 ( n − 1 ) ] \displaystyle[\frac{(n-1)s^2}{\chi_{\frac{\alpha}{2}}^{2}(n-1)},\frac{(n-1)s^2}{\chi_{1-\frac{\alpha}{2}}^{2}(n-1)}] [χ2α2(n1)(n1)s2,χ12α2(n1)(n1)s2]

总结

你可能感兴趣的:(概率论与数理统计,概率论,数学)