经验分布弱收敛和一致收敛到某一特定分布
定理:
若
F
n
⇒
F
F_n \Rightarrow F
Fn⇒F, 且
F
F
F 连续,则:
lim
n
→
∞
sup
t
∣
F
n
(
t
)
−
F
(
t
)
∣
=
0
\lim\limits_{n \to \infty} \sup_{t} | F_n(t) - F(t)| =0
n→∞limtsup∣Fn(t)−F(t)∣=0
Proof:
由于
F
F
F 是某一分布函数,所以必定
∃
M
>
0
\exist M>0
∃M>0 ,使得对于
∀
ε
>
0
\forall \varepsilon >0
∀ε>0,有:
{
1
−
F
(
M
)
<
ε
F
(
−
M
)
<
ε
\left\{ \begin{matrix} 1- F(M) < \varepsilon \\ F(-M) < \varepsilon \end{matrix} \right.
{1−F(M)<εF(−M)<ε
对于区间
[
−
M
,
M
]
[-M,M]
[−M,M],一定存在划分
−
M
=
x
0
<
x
1
<
⋯
<
x
J
=
M
-M = x_0 < x_1< \cdots < x_J = M
−M=x0<x1<⋯<xJ=M,使得:
对于上述的
ε
>
0
\varepsilon >0
ε>0,
∃
δ
>
0
\exist\ \delta >0
∃ δ>0,当
x
i
+
1
−
x
i
<
δ
x_{i+1} -x_i < \delta
xi+1−xi<δ 时,
F
(
x
i
+
1
)
−
F
(
x
i
)
<
ε
(1)
F(x_{i+1}) - F(x_i) < \varepsilon\tag{1}
F(xi+1)−F(xi)<ε(1).
(这里其实是给出了每个小区间内函数差值上界)
已知
F
n
⇒
F
F_n \Rightarrow F
Fn⇒F,那么,对于上述的
ε
>
0
\varepsilon >0
ε>0,
∃
N
∈
N
\exist \ N \in \mathbb{N}
∃ N∈N,使得当
n
>
N
n > N
n>N时:
∣
F
n
(
x
i
)
−
F
(
x
)
∣
≤
ε
(2)
|F_n(x_i) - F(x)| \leq \varepsilon\tag{2}
∣Fn(xi)−F(x)∣≤ε(2)
(这里给出了每个小区间内
F
n
F_n
Fn和
F
F
F的差值上界)
故, ∀ x ∈ [ − M , M ] \forall x \in [-M,M] ∀x∈[−M,M], ∃ i → x ∈ [ x i , x i + 1 ] \exist i \rightarrow x \in [x_i,x_{i+1}] ∃i→x∈[xi,xi+1] ,此时:
∣
F
n
(
x
)
−
F
(
x
)
∣
≤
∣
F
n
(
x
i
+
1
)
−
F
(
x
)
∣
(
因为
F
n
单调不减
)
=
∣
F
n
(
x
i
+
1
)
−
F
(
x
i
+
1
)
+
F
(
x
i
+
1
)
−
F
(
x
)
∣
≤
∣
F
n
(
x
i
+
1
)
−
F
(
x
i
+
1
)
∣
+
∣
F
(
x
i
+
1
)
−
F
(
x
)
∣
(
三角不等式
)
≤
2
ε
(
用上面的
(
1
)
(
2
)
)
\begin{aligned} |F_n(x) - F(x)| & \leq |F_n(x_{i+1}) - F(x)|(因为F_n单调不减)\\ & = |F_n(x_{i+1}) - F(x_{i+1})+F(x_{i+1})-F(x)|\\ & \leq |F_n(x_{i+1}) - F(x_{i+1})| + |F(x_{i+1})-F(x)|(三角不等式)\\ & \leq 2\varepsilon(用上面的(1)(2)) \end{aligned}
∣Fn(x)−F(x)∣≤∣Fn(xi+1)−F(x)∣(因为Fn单调不减)=∣Fn(xi+1)−F(xi+1)+F(xi+1)−F(x)∣≤∣Fn(xi+1)−F(xi+1)∣+∣F(xi+1)−F(x)∣(三角不等式)≤2ε(用上面的(1)(2))
由于
ε
\varepsilon
ε 是任意的,可以任意小,所以有:
lim
n
→
∞
sup
t
∣
F
n
(
t
)
−
F
(
t
)
∣
=
0
\lim\limits_{n \to \infty} \sup_{t} | F_n(t) - F(t)| =0
n→∞limtsup∣Fn(t)−F(t)∣=0
得证。