矩估计法
用样本矩估计相应的总体矩,用样本矩的函数估计总体矩相应的函数,然后求出要估计的参数,称这种估计法为矩估计法
步骤
设
X
X
X为连续型随机变量,其概率密度为
f
(
x
;
θ
1
,
θ
2
,
⋯
,
θ
k
)
f(x;\theta_{1},\theta_{2},\cdots,\theta_{k})
f(x;θ1,θ2,⋯,θk),或
X
X
X为离散型随机变量,其分布律为
P
{
X
=
x
}
=
p
(
x
;
θ
1
,
θ
2
,
⋯
,
θ
k
)
P \left\{X=x\right\}=p(x;\theta_{1},\theta_{2},\cdots,\theta_{k})
P{X=x}=p(x;θ1,θ2,⋯,θk),其中
θ
1
,
θ
2
,
⋯
,
θ
k
\theta_{1},\theta_{2},\cdots,\theta_{k}
θ1,θ2,⋯,θk为待估计参数,
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn是来自
X
X
X的样本。假设总体
X
X
X的前
k
k
k阶矩
μ
l
=
E
(
X
l
)
=
∫
−
∞
+
∞
x
l
f
(
x
;
θ
1
,
θ
2
,
⋯
,
θ
k
)
d
x
(
X
连续型
)
μ
l
=
E
(
X
l
)
=
∑
x
∈
R
X
x
l
p
(
x
;
θ
1
,
θ
2
,
⋯
,
θ
k
)
(
X
离散型
)
\begin{aligned} \mu_{l}=E(X^{l})&=\int_{-\infty}^{+\infty}x^{l}f(x;\theta_{1},\theta_{2},\cdots,\theta_{k})dx \quad (X连续型)\\ \mu_{l}=E(X^{l})&=\sum\limits_{x \in R_{X}}^{}x^{l}p(x;\theta_{1},\theta_{2},\cdots,\theta_{k})\quad (X离散型) \end{aligned}
μl=E(Xl)μl=E(Xl)=∫−∞+∞xlf(x;θ1,θ2,⋯,θk)dx(X连续型)=x∈RX∑xlp(x;θ1,θ2,⋯,θk)(X离散型)
(其中
R
X
R_{X}
RX是
X
X
X可能取值的范围)存在。一般来说,它们是
θ
1
,
θ
2
,
⋯
,
θ
k
\theta_{1},\theta_{2},\cdots,\theta_{k}
θ1,θ2,⋯,θk的函数。基于样本矩
A
l
=
1
n
∑
i
=
1
n
X
i
l
A_{l}=\frac{1}{n}\sum\limits_{i=1}^{n}X_{i}^{l}
Al=n1i=1∑nXil
依概率收敛于相应的总体距
μ
l
(
l
=
1
,
2
,
⋯
,
k
)
\mu_{l}(l=1,2,\cdots ,k)
μl(l=1,2,⋯,k),样本矩的连续函数依概率收敛于相应的总体矩的连续函数,我们就用样本矩作为相应的总体矩的估计量,而以样本矩的连续函数作为相应的总体矩的连续函数的估计量。这种估计方法称为矩估计法。矩估计法的具体做法如下:设
{
μ
1
=
μ
1
(
θ
1
,
θ
2
,
⋯
,
θ
k
)
μ
2
=
μ
2
(
θ
1
,
θ
2
,
⋯
,
θ
k
)
⋮
μ
k
=
μ
k
(
θ
1
,
θ
2
,
⋯
,
θ
k
)
\left\{\begin{aligned}&\mu_{1}=\mu_{1}(\theta_{1},\theta_{2},\cdots,\theta_{k})\\ &\mu_{2}=\mu_{2}(\theta_{1},\theta_{2},\cdots,\theta_{k})\\ &\vdots \\&\mu_{k}=\mu_{k}(\theta_{1},\theta_{2},\cdots,\theta_{k})\end{aligned}\right.
⎩
⎨
⎧μ1=μ1(θ1,θ2,⋯,θk)μ2=μ2(θ1,θ2,⋯,θk)⋮μk=μk(θ1,θ2,⋯,θk)
这是一个包含
k
k
k个未知参数
θ
1
,
θ
2
,
⋯
,
θ
k
\theta_{1},\theta_{2},\cdots,\theta_{k}
θ1,θ2,⋯,θk的联立方程组。一般来说,可以从中解出
θ
1
,
θ
2
,
⋯
,
θ
k
\theta_{1},\theta_{2},\cdots,\theta_{k}
θ1,θ2,⋯,θk,得到
{
θ
1
=
θ
1
(
μ
1
,
μ
2
,
⋯
,
μ
k
)
θ
2
=
θ
2
(
μ
1
,
μ
2
,
⋯
,
μ
k
)
⋮
θ
k
=
θ
k
(
μ
1
,
μ
2
,
⋯
,
μ
k
)
\left\{\begin{aligned}&\theta_{1}=\theta_{1}(\mu_{1},\mu_{2},\cdots ,\mu_{k})\\&\theta_{2}=\theta_{2}(\mu_{1},\mu_{2},\cdots ,\mu_{k})\\ \vdots \\&\theta_{k}=\theta_{k}(\mu_{1},\mu_{2},\cdots ,\mu_{k})\end{aligned}\right.
⎩
⎨
⎧⋮θ1=θ1(μ1,μ2,⋯,μk)θ2=θ2(μ1,μ2,⋯,μk)θk=θk(μ1,μ2,⋯,μk)
以
A
i
A_{i}
Ai分别代替上式中的
μ
i
,
i
=
1
,
2
,
⋯
,
k
\mu_{i},i=1,2,\cdots ,k
μi,i=1,2,⋯,k,就以
θ
i
^
=
θ
i
(
A
1
,
A
2
,
⋯
,
A
k
)
,
i
=
1
,
2
,
⋯
,
k
\hat{\theta_{i}}=\theta_{i}(A_{1},A_{2},\cdots ,A_{k}),i=1,2,\cdots ,k
θi^=θi(A1,A2,⋯,Ak),i=1,2,⋯,k
分别作为
θ
i
,
i
=
1
,
2
,
⋯
,
k
\theta_{i},i=1,2,\cdots ,k
θi,i=1,2,⋯,k的估计量,这种估计量称为矩估计量。矩估计量的观察值称为矩估计值
例1:设总体 X X X在 [ a , b ] [a,b] [a,b]上服从均匀分布, a , b a,b a,b未知, X 1 , X 2 , ⋯ , X n X_{1},X_{2},\cdots,X_{n} X1,X2,⋯,Xn是来自 X X X的样本,试求 a , b a,b a,b的矩估计量
μ
1
=
E
(
X
)
=
a
+
b
2
μ
2
=
E
(
X
2
)
=
D
(
X
)
+
[
E
(
X
)
]
2
=
(
b
−
a
)
2
12
+
(
a
+
b
)
2
4
⇒
{
a
+
b
=
2
μ
1
b
−
a
=
12
(
μ
2
−
μ
1
2
)
⇒
{
a
=
μ
1
−
3
(
μ
2
−
μ
1
2
)
b
=
μ
1
+
3
(
μ
2
−
μ
1
2
)
\begin{aligned} \mu_{1}&=E(X)=\frac{a+b}{2}\\ \mu_{2}&=E(X^{2})=D(X)+[E(X)]^{2}=\frac{(b-a)^{2}}{12}+ \frac{(a+b)^{2}}{4}\\ &\Rightarrow \left\{\begin{aligned}&a+b=2\mu_{1}\\&b-a=\sqrt{12(\mu_{2}-\mu_{1}^{2})}\end{aligned}\right.\\ &\Rightarrow \left\{\begin{aligned}&a=\mu_{1}-\sqrt{3(\mu_{2}-\mu_{1}^{2})}\\&b=\mu_{1}+\sqrt{3(\mu_{2}-\mu_{1}^{2})}\end{aligned}\right. \end{aligned}
μ1μ2=E(X)=2a+b=E(X2)=D(X)+[E(X)]2=12(b−a)2+4(a+b)2⇒⎩
⎨
⎧a+b=2μ1b−a=12(μ2−μ12)⇒⎩
⎨
⎧a=μ1−3(μ2−μ12)b=μ1+3(μ2−μ12)
注意到
1
n
∑
i
=
1
n
X
i
2
−
X
ˉ
2
=
1
n
∑
i
=
1
n
(
X
i
−
X
ˉ
)
2
\begin{aligned} \frac{1}{n}\sum\limits_{i=1}^{n}X_{i}^{2}-\bar{X}^{2}=\frac{1}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2}\end{aligned}
n1i=1∑nXi2−Xˉ2=n1i=1∑n(Xi−Xˉ)2,分别以
A
1
,
A
2
A_{1},A_{2}
A1,A2代替
μ
1
,
μ
2
\mu_{1},\mu_{2}
μ1,μ2,得到
a
,
b
a,b
a,b的矩估计量分别为
a
^
=
A
1
−
3
(
A
2
−
A
1
2
)
=
X
ˉ
−
3
n
∑
i
=
1
n
(
X
i
−
X
ˉ
)
2
b
^
=
A
1
+
3
(
A
2
−
A
1
2
)
=
X
ˉ
+
3
n
∑
i
=
1
n
(
X
i
−
X
ˉ
)
2
\begin{aligned} \hat{a}&=A_{1}-\sqrt{3(A_{2}-A_{1}^{2})}=\bar{X}-\sqrt{\frac{3}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2}}\\ \hat{b}&=A_{1}+\sqrt{3(A_{2}-A_{1}^{2})}=\bar{X}+\sqrt{\frac{3}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2}} \end{aligned}
a^b^=A1−3(A2−A12)=Xˉ−n3i=1∑n(Xi−Xˉ)2=A1+3(A2−A12)=Xˉ+n3i=1∑n(Xi−Xˉ)2
例2:设总体 X X X的均值 μ \mu μ及方差 σ 2 \sigma^{2} σ2都存在,且有 σ 2 > 0 \sigma^{2}>0 σ2>0。但 μ , σ 2 \mu,\sigma^{2} μ,σ2均为未知,又设 X 1 , X 2 , ⋯ , X n X_{1},X_{2},\cdots,X_{n} X1,X2,⋯,Xn是来自 X X X得样本。试求 μ , σ 2 \mu,\sigma^{2} μ,σ2的矩估计量
μ
1
=
E
(
X
)
=
μ
μ
2
=
E
(
X
2
)
=
D
(
X
)
+
[
E
(
X
)
]
2
=
σ
2
+
μ
⇒
{
μ
=
μ
1
σ
2
=
μ
2
−
μ
1
2
\begin{aligned} \mu_{1}&=E(X)=\mu\\ \mu_{2}&=E(X^{2})=D(X)+[E(X)]^{2}=\sigma^{2}+\mu\\ &\Rightarrow \left\{\begin{aligned}&\mu=\mu_{1}\\&\sigma^{2}=\mu_{2}-\mu_{1}^{2}\end{aligned}\right. \end{aligned}
μ1μ2=E(X)=μ=E(X2)=D(X)+[E(X)]2=σ2+μ⇒{μ=μ1σ2=μ2−μ12
分别以
A
1
,
A
2
A_{1},A_{2}
A1,A2代替
μ
1
,
μ
2
\mu_{1},\mu_{2}
μ1,μ2,得
μ
\mu
μ和
σ
2
\sigma^{2}
σ2的矩估计量分别为
μ
^
=
A
1
=
X
ˉ
σ
2
^
=
A
2
−
A
1
2
=
1
n
∑
i
=
1
n
X
i
2
−
X
ˉ
2
=
1
n
∑
i
=
1
n
(
X
i
−
X
ˉ
)
2
\begin{aligned} \hat{\mu}&=A_{1}=\bar{X}\\ \hat{\sigma^{2}}&=A_{2}-A_{1}^{2}=\frac{1}{n}\sum\limits_{i=1}^{n}X_{i}^{2}-\bar{X}^{2}=\frac{1}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2} \end{aligned}
μ^σ2^=A1=Xˉ=A2−A12=n1i=1∑nXi2−Xˉ2=n1i=1∑n(Xi−Xˉ)2
所得结果表名,总体均值与方差的矩估计量的表达式不因不同的总体分布而异
最大似然估计
若总体
X
X
X属离散型,其分布律
P
{
X
=
x
}
=
p
{
x
;
θ
}
,
θ
∈
Θ
P \left\{X=x\right\}=p \left\{x;\theta\right\},\theta \in \Theta
P{X=x}=p{x;θ},θ∈Θ的形式已知,
θ
\theta
θ为待估参数,
Θ
\Theta
Θ是
θ
\theta
θ可能取值的范围。设
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn是来自
X
X
X的样本,则
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn的联合分布律为
∏
i
=
1
n
p
(
x
i
;
θ
)
\prod\limits_{i=1}^{n}p(x_{i};\theta)
i=1∏np(xi;θ)
又设
x
1
,
x
2
,
⋯
,
x
n
x_{1},x_{2},\cdots,x_{n}
x1,x2,⋯,xn是相应于样本
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn的一个样本值。易知样本
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn取到观察值
x
1
,
x
2
,
⋯
,
x
n
x_{1},x_{2},\cdots,x_{n}
x1,x2,⋯,xn的概率,亦即事件
{
X
1
=
x
1
,
X
2
=
x
2
,
⋯
,
X
n
,
x
n
}
\left\{X_{1}=x_{1},X_{2}=x_{2},\cdots ,X_{n},x_{n}\right\}
{X1=x1,X2=x2,⋯,Xn,xn}发生的概率为
L
(
θ
)
=
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
)
=
∏
i
=
1
n
p
(
x
i
;
θ
)
,
θ
∈
Θ
L(\theta)=L(x_{1},x_{2},\cdots,x_{n};\theta)=\prod\limits_{i=1}^{n}p(x_{i};\theta),\theta \in \Theta
L(θ)=L(x1,x2,⋯,xn;θ)=i=1∏np(xi;θ),θ∈Θ
这一概率随
θ
\theta
θ的取值而变化,它是
θ
\theta
θ的函数,
L
(
θ
)
L(\theta)
L(θ)称为样本的似然函数(注意,这里
x
1
,
x
2
,
⋯
,
x
n
x_{1},x_{2},\cdots,x_{n}
x1,x2,⋯,xn是已知的样本值,它们都是常数
在
θ
\theta
θ取值的可能范围
Θ
\Theta
Θ内挑选使似然函数
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
)
L(x_{1},x_{2},\cdots,x_{n};\theta)
L(x1,x2,⋯,xn;θ)达到最大的参数值
θ
^
\hat{\theta}
θ^,作为参数
θ
\theta
θ的估计值。即使
θ
^
\hat{\theta}
θ^使
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
^
)
=
max
θ
∈
Θ
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
)
L(x_{1},x_{2},\cdots,x_{n};\hat{\theta})=\max\limits_{\theta \in \Theta}L(x_{1},x_{2},\cdots,x_{n};\theta)
L(x1,x2,⋯,xn;θ^)=θ∈ΘmaxL(x1,x2,⋯,xn;θ)
这样得到的
θ
^
\hat{\theta}
θ^与样本值
x
1
,
x
2
,
⋯
,
x
n
x_{1},x_{2},\cdots,x_{n}
x1,x2,⋯,xn有关,常记为
θ
^
(
x
1
,
x
2
,
⋯
,
x
n
)
\hat{\theta}(x_{1},x_{2},\cdots,x_{n})
θ^(x1,x2,⋯,xn),称为参数
θ
\theta
θ的最大似然估计值,而相应的统计量
θ
^
(
X
1
,
X
2
,
⋯
,
X
n
)
\hat{\theta}(X_{1},X_{2},\cdots,X_{n})
θ^(X1,X2,⋯,Xn)称为参数
θ
\theta
θ的最大似然估计量
若总体
X
X
X属连续型,其概率密度
f
(
x
;
θ
)
,
θ
∈
Θ
f(x;\theta),\theta \in \Theta
f(x;θ),θ∈Θ的形式已知,
θ
\theta
θ为待估参数,
Θ
\Theta
Θ是
θ
\theta
θ可能取值的范围,设
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn是来自
X
X
X得样本,则
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn的联合密度为
∏
i
=
1
n
f
(
x
i
,
θ
)
\prod\limits_{i=1}^{n}f(x_{i},\theta)
i=1∏nf(xi,θ)
设
x
1
,
x
2
,
⋯
,
x
n
x_{1},x_{2},\cdots,x_{n}
x1,x2,⋯,xn是相应于样本
X
1
,
X
2
,
⋯
,
X
n
X_{1},X_{2},\cdots,X_{n}
X1,X2,⋯,Xn的一个样本值,则随机点
(
X
1
,
X
2
,
⋯
,
X
n
)
(X_{1},X_{2},\cdots,X_{n})
(X1,X2,⋯,Xn)落在点
(
x
1
,
x
2
,
⋯
,
x
n
)
(x_{1},x_{2},\cdots,x_{n})
(x1,x2,⋯,xn)的邻域(边长分别为
d
x
1
,
d
x
2
,
⋯
,
d
x
n
dx_{1},d x_{2} ,\cdots ,d x_{n}
dx1,dx2,⋯,dxn的
n
n
n维立方体)内的概率近似地为
∏
i
=
1
n
f
(
x
i
;
θ
)
d
x
\prod\limits_{i=1}^{n}f(x_{i};\theta)dx
i=1∏nf(xi;θ)dx
其值随
θ
\theta
θ的取值而变化,与离散型的情况一样,我们取
θ
\theta
θ的估计量
θ
^
\hat{\theta}
θ^使概率取到最大值,但因子
∏
i
=
1
n
d
x
i
\prod\limits_{i=1}^{n}dx_{i}
i=1∏ndxi不随
θ
\theta
θ而变,故只需考虑函数
L
(
θ
)
=
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
)
=
∏
i
=
1
n
f
(
x
i
;
θ
)
L(\theta)=L(x_{1},x_{2},\cdots,x_{n} ;\theta)=\prod\limits_{i=1}^{n}f(x_{i};\theta)
L(θ)=L(x1,x2,⋯,xn;θ)=i=1∏nf(xi;θ)
的最大值,这里
L
(
θ
)
L(\theta)
L(θ)称为样本的似然函数,若
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
)
=
max
θ
∈
Θ
L
(
x
1
,
x
2
,
⋯
,
x
n
;
θ
)
L(x_{1},x_{2},\cdots,x_{n};\theta)=\max\limits_{\theta \in \Theta}L(x_{1},x_{2},\cdots,x_{n};\theta)
L(x1,x2,⋯,xn;θ)=θ∈ΘmaxL(x1,x2,⋯,xn;θ)
则称
θ
^
(
x
1
,
x
2
,
⋯
,
x
n
)
\hat{\theta}(x_{1},x_{2},\cdots,x_{n})
θ^(x1,x2,⋯,xn)为
θ
\theta
θ的最大似然估计值,称
θ
^
(
X
1
,
X
2
,
⋯
,
X
n
)
\hat{\theta}(X_{1},X_{2},\cdots,X_{n})
θ^(X1,X2,⋯,Xn)为
θ
\theta
θ的最大似然估计量
这样,确定最大似然估计量的问题就归结为微分学中秋最大值的问题了,
在很多情形下,
p
(
x
;
θ
)
p(x;\theta)
p(x;θ)和
f
(
x
;
θ
)
f(x;\theta)
f(x;θ)关于
θ
\theta
θ可微,这时
θ
^
\hat{\theta}
θ^常可从方程
d
d
θ
L
(
θ
)
=
0
\frac{d}{d \theta}L(\theta)=0
dθdL(θ)=0
解得。又因
L
(
θ
)
L(\theta)
L(θ)与
L
(
ln
θ
)
L(\ln \theta)
L(lnθ)在同一
θ
\theta
θ处取到极值,因此,
θ
\theta
θ的最大似然估计
θ
\theta
θ也可以从方程
d
d
θ
ln
L
(
θ
)
=
0
\frac{d}{d \theta}\ln L(\theta)=0
dθdlnL(θ)=0
求得,称为对数似然方程
例3:设 X ∼ N ( μ , σ 2 ) X \sim N(\mu,\sigma^{2}) X∼N(μ,σ2),其中 μ , σ 2 ( σ > 0 ) \mu,\sigma^{2}(\sigma>0) μ,σ2(σ>0)均为未知参数,从总体取得样本为 X 1 , X 2 , ⋯ , X n X_{1},X_{2},\cdots,X_{n} X1,X2,⋯,Xn,试求
- μ \mu μ和 σ 2 \sigma^{2} σ2的矩估计
E ( X ) = μ E ( X 2 ) = D ( X ) + [ E ( X ) ] 2 = σ 2 + μ 2 ⇒ { μ = X ˉ σ 2 + μ 2 = 1 n ∑ i = 1 n X i 2 ⇒ { μ ^ = X ˉ σ ^ = 1 n ∑ i = 1 n X i 2 − X ˉ 2 = 1 n ∑ i = 1 n ( X i − X ˉ ) 2 \begin{aligned} E(X)&=\mu\\ E(X^{2})&=D(X)+[E(X)]^{2}=\sigma^{2}+\mu^{2}\\ & \Rightarrow \left\{\begin{aligned}&\mu=\bar{X}\\&\sigma^{2}+\mu^{2}=\frac{1}{n}\sum\limits_{i=1}^{n}X_{i}^{2}\end{aligned}\right.\\ & \Rightarrow \left\{\begin{aligned}&\hat{\mu}=\bar{X}\\&\hat{\sigma}=\frac{1}{n}\sum\limits_{i=1}^{n}X_{i}^{2}-\bar{X}^{2}=\frac{1}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2}\end{aligned}\right. \end{aligned} E(X)E(X2)=μ=D(X)+[E(X)]2=σ2+μ2⇒⎩ ⎨ ⎧μ=Xˉσ2+μ2=n1i=1∑nXi2⇒⎩ ⎨ ⎧μ^=Xˉσ^=n1i=1∑nXi2−Xˉ2=n1i=1∑n(Xi−Xˉ)2
二阶矩可以用中心矩直接得到 σ ^ = 1 n ∑ i = 1 n X i 2 − X ˉ 2 = 1 n ∑ i = 1 n ( X i − X ˉ ) 2 \begin{aligned} \hat{\sigma}=\frac{1}{n}\sum\limits_{i=1}^{n}X_{i}^{2}-\bar{X}^{2}=\frac{1}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2}\end{aligned} σ^=n1i=1∑nXi2−Xˉ2=n1i=1∑n(Xi−Xˉ)2
- μ \mu μ和 σ 2 \sigma^{2} σ2的最大似然估计
由于
X
X
X的密度为
1
2
π
σ
e
−
(
x
−
μ
)
2
2
σ
2
,
−
∞
<
x
<
+
∞
\begin{aligned} \frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{(x-\mu)^{2}}{2\sigma^{2}}},-\infty<x<+\infty\end{aligned}
2πσ1e−2σ2(x−μ)2,−∞<x<+∞,所以
L
(
μ
,
σ
2
)
=
(
1
2
π
σ
)
n
e
−
1
2
σ
2
∑
i
=
1
n
(
X
i
−
μ
)
2
ln
L
(
μ
,
σ
2
)
=
−
n
2
ln
(
2
π
)
−
n
2
ln
(
σ
2
)
−
1
2
σ
2
∑
i
=
1
n
(
X
i
−
μ
)
2
⇒
{
∂
ln
L
∂
μ
=
1
σ
2
∑
i
=
1
n
(
X
i
−
μ
)
=
0
∂
ln
L
∂
σ
2
=
−
n
2
1
σ
2
+
1
2
(
σ
2
)
2
∑
i
=
1
n
(
X
i
−
μ
)
2
=
0
⇒
{
μ
^
−
X
ˉ
σ
2
^
=
1
n
∑
i
=
1
n
(
X
i
−
X
ˉ
)
2
\begin{aligned} L(\mu,\sigma^{2})&=\left(\frac{1}{\sqrt{2\pi}\sigma}\right)^{n}e^{-\frac{1}{2\sigma^{2}}\sum\limits_{i=1}^{n}(X_{i}-\mu)^{2}}\\ \ln L(\mu,\sigma^{2})&=- \frac{n}{2}\ln (2\pi)- \frac{n}{2}\ln (\sigma^{2})- \frac{1}{2\sigma^{2}}\sum\limits_{i=1}^{n}(X_{i}-\mu)^{2}\\ &\Rightarrow \left\{\begin{aligned}&\frac{\partial \ln L}{\partial \mu}=\frac{1}{\sigma^{2}}\sum\limits_{i=1}^{n}(X_{i}-\mu)=0\\&\frac{\partial \ln L}{\partial \sigma^{2}}=- \frac{n}{2} \frac{1}{\sigma^{2}}+\frac{1}{2(\sigma^{2})^{2}}\sum\limits_{i=1}^{n}(X_{i}-\mu)^{2}=0\end{aligned}\right.\\ &\Rightarrow \left\{\begin{aligned}&\hat{\mu}-\bar{X}\\&\hat{\sigma^{2}}=\frac{1}{n}\sum\limits_{i=1}^{n}(X_{i}-\bar{X})^{2}\end{aligned}\right. \end{aligned}
L(μ,σ2)lnL(μ,σ2)=(2πσ1)ne−2σ21i=1∑n(Xi−μ)2=−2nln(2π)−2nln(σ2)−2σ21i=1∑n(Xi−μ)2⇒⎩
⎨
⎧∂μ∂lnL=σ21i=1∑n(Xi−μ)=0∂σ2∂lnL=−2nσ21+2(σ2)21i=1∑n(Xi−μ)2=0⇒⎩
⎨
⎧μ^−Xˉσ2^=n1i=1∑n(Xi−Xˉ)2
例4:设总体 X ∼ U [ − θ , θ ] , X 1 , X 2 , ⋯ , X n X \sim U[-\theta,\theta],X_{1},X_{2},\cdots,X_{n} X∼U[−θ,θ],X1,X2,⋯,Xn是来自总体 X X X的简单随机样本,试求参数 θ \theta θ的最大似然估计
似然函数为
L
(
θ
)
=
∏
i
=
1
n
f
(
x
i
)
=
{
1
(
2
θ
)
n
−
θ
≤
x
1
,
x
2
,
⋯
,
x
n
≤
θ
0
其他
L(\theta)=\prod\limits_{i=1}^{n}f(x_{i})=\left\{\begin{aligned}& \frac{1}{(2\theta)^{n}}&-\theta \leq x_{1},x_{2},\cdots,x_{n}\leq \theta\\&0&其他\end{aligned}\right.
L(θ)=i=1∏nf(xi)=⎩
⎨
⎧(2θ)n10−θ≤x1,x2,⋯,xn≤θ其他
这里求导只能得到最小值
显然,
θ
\theta
θ越小,
1
(
2
θ
)
n
\begin{aligned} \frac{1}{(2\theta)^{n}}\end{aligned}
(2θ)n1越大,但
θ
\theta
θ必须满足
−
θ
≤
x
1
,
x
2
,
⋯
,
x
n
≤
θ
-\theta \leq x_{1},x_{2},\cdots,x_{n}\leq \theta
−θ≤x1,x2,⋯,xn≤θ,也就是必有
{
θ
≥
max
(
x
1
,
x
2
,
⋯
,
x
n
)
−
θ
≤
min
(
x
1
,
x
2
,
⋯
,
x
n
)
⇒
θ
≥
max
(
∣
x
1
∣
,
∣
x
2
∣
,
⋯
,
∣
x
n
∣
)
\begin{aligned} &\left\{\begin{aligned}&\theta \geq \max (x_{1},x_{2},\cdots,x_{n})\\&-\theta \leq \min (x_{1},x_{2},\cdots,x_{n})\end{aligned}\right.\\ & \Rightarrow \theta \geq \max (|x_{1}|,|x_{2}|,\cdots ,|x_{n}|) \end{aligned}
{θ≥max(x1,x2,⋯,xn)−θ≤min(x1,x2,⋯,xn)⇒θ≥max(∣x1∣,∣x2∣,⋯,∣xn∣)
即有
θ
\theta
θ的最大似然估计
θ
^
=
max
(
∣
X
1
∣
,
∣
X
2
∣
,
⋯
,
∣
X
n
∣
)
\hat{\theta}=\max (|X_{1}|,|X_{2}|,\cdots ,|X_{n}|)
θ^=max(∣X1∣,∣X2∣,⋯,∣Xn∣)
CSDN话题挑战赛第2期
参赛话题:学习笔记