随机变量
1. 随机变量
随机变量: 定义在样本空间上的实值函数,称为随机变量。例如掷骰子时,我们关心的是两颗骰子的点数和,而不是两次骰子具体的点数,我们所关注的量,即点数和,就是随机变量。
示性随机变量:
I
=
{
1
电
池
的
寿
命
是
两
年
或
者
更
长
0
其
他
情
形
I = \begin{cases} 1 & 电池的寿命是两年或者更长 \\ 0 & 其他情形 \end{cases}
I={10电池的寿命是两年或者更长其他情形
以E记电池的寿命是两年或者更长,那么随机变量
I
I
I称为事件E的示性随机变量。
I
I
I的取值依赖于E是否发生
概率质量函数: 描述离散型随机变量的输出值
概率密度函数: 描述连续型随机变量的输出值
累积分布函数F:
∀
a
∈
R
,
F
(
a
)
=
P
{
X
≤
a
}
=
{
∑
∀
x
i
≤
a
p
(
x
i
)
X
为
离
散
随
机
变
量
∫
−
∞
a
f
(
x
)
d
x
X
为
连
续
随
机
变
量
\forall a \in R,F(a) = P\{X \leq a \} = \begin{cases} \sum \limits_{\forall x_i \leq a} p(x_i) & X为离散随机变量 \\ \int \limits_{- \infty}^a f(x) dx & X为连续随机变量 \end{cases}
∀a∈R,F(a)=P{X≤a}=⎩⎪⎨⎪⎧∀xi≤a∑p(xi)−∞∫af(x)dxX为离散随机变量X为连续随机变量
F的性质:
1. F(b)是b的非减函数
2.
lim
b
→
∞
F
(
b
)
=
F
(
∞
)
=
1
\lim_{b \to \infty}F(b) = F(\infty) = 1
limb→∞F(b)=F(∞)=1
3.
lim
b
→
−
∞
F
(
b
)
=
F
(
−
∞
)
=
0
\lim_{b \to -\infty} F(b) = F(-\infty) = 0
limb→−∞F(b)=F(−∞)=0
∀ a < b , 有 P { a < X ≤ b } = F ( b ) − F ( a ) \forall a < b,有P\{a < X \leq b\} = F(b) - F(a) ∀a<b,有P{a<X≤b}=F(b)−F(a)
P { X < b } = lim h → 0 + P { X ≤ b − h } = lim h → 0 + F ( b − h ) P\{X < b\} = \lim \limits_{h \to 0^+} P\{X \leq b - h\} = \lim \limits_{h \to 0^+} F(b - h) P{X<b}=h→0+limP{X≤b−h}=h→0+limF(b−h)
P { X < b } P\{X < b\} P{X<b}不一定等于F(b),因为F(b)也包括X=b的概率
变量期望E(X):
E
(
x
)
=
{
∑
k
∞
x
p
(
x
)
x
为
离
散
随
机
变
量
∫
−
∞
∞
x
f
(
x
)
d
x
x
为
连
续
随
机
变
量
E(x) = \begin{cases} \sum \limits^{\infty}_{k}x p(x) & x为离散随机变量 \\ \int \limits^{\infty}_{- \infty} xf(x) dx & x为连续随机变量 \end{cases}
E(x)=⎩⎪⎪⎨⎪⎪⎧k∑∞xp(x)−∞∫∞xf(x)dxx为离散随机变量x为连续随机变量
函数期望E[g(X)]:
E
[
g
(
x
)
]
=
{
∑
k
∞
g
(
x
)
p
(
x
)
x
为
离
散
随
机
变
量
∫
−
∞
∞
g
(
x
)
f
(
x
)
d
x
x
为
连
续
随
机
变
量
E[g(x)] = \begin{cases} \sum \limits^{\infty}_{k} g(x) p(x) & x为离散随机变量 \\ \int \limits^{\infty}_{- \infty} g(x) f(x) dx & x为连续随机变量 \end{cases}
E[g(x)]=⎩⎪⎪⎨⎪⎪⎧k∑∞g(x)p(x)−∞∫∞g(x)f(x)dxx为离散随机变量x为连续随机变量
若a和b都是常数,则
E
[
a
X
+
b
]
=
a
E
[
X
]
+
b
E[aX + b] = aE[X] + b
E[aX+b]=aE[X]+b
随机变量X的期望E(X)称为均值或X的一阶矩。
E
(
X
n
)
,
n
≥
1
E(X^n),\ n \geq 1
E(Xn), n≥1称为X的n阶矩。
E [ X n ] = { ∑ x : p ( x ) > 0 x n p ( x ) X 为 离 散 ∫ − ∞ ∞ x n f ( x ) d x X 为 连 续 E[X^n] = \begin{cases} \sum \limits_{x:p(x) > 0} x^n p(x) & X为离散 \\ \\ \int \limits^\infty_{- \infty} x^n f(x) dx & X 为连续 \end{cases} E[Xn]=⎩⎪⎪⎪⎪⎨⎪⎪⎪⎪⎧x:p(x)>0∑xnp(x)−∞∫∞xnf(x)dxX为离散X为连续
方差D(X):
D
(
X
)
=
E
[
(
X
−
E
[
X
]
)
2
]
=
E
(
X
2
)
−
[
E
(
X
)
]
2
D(X) = E[(X - E[X])^2] = E(X^2) - [E(X)]^2
D(X)=E[(X−E[X])2]=E(X2)−[E(X)]2
例如:
X代表掷一颗均匀的骰子的结果,求Var(X)
解:
E
(
X
)
=
∑
i
=
1
6
i
×
1
6
=
7
2
E
(
X
2
)
=
∑
i
=
1
6
i
2
×
1
6
=
91
6
V
a
r
(
X
)
=
91
6
−
(
7
2
)
2
=
35
12
E(X) = \sum \limits^6_{i = 1} i \times \frac{1}{6} = \frac{7}{2} \\ E(X^2) = \sum \limits^6_{i = 1} i^2 \times \frac{1}{6} = \frac{91}{6} \\ Var(X) = \frac{91}{6} - (\frac{7}{2})^ 2 = \frac{35}{12}
E(X)=i=1∑6i×61=27E(X2)=i=1∑6i2×61=691Var(X)=691−(27)2=1235
易混淆概念:
1. 概率质量函数针对离散随机变量
2. 概率密度函数针对连续随机变量
3. 累积分布函数也叫分布函数,是概率密度函数的积分
2. 离散随机变量、期望、方差
2.1 离散随机变量
一个最多取可数个可能值的随机变量,称为离散随机变量。对于一个离散随机变量X,用 p ( a ) = P { X = a } p(a) = P\{X = a\} p(a)=P{X=a}定义概率质量函数p(a)
X ∈ { x 1 , x 2 , ⋯ } ⇒ { p ( x i ) > 0 i = 1 , 2 , ⋯ p ( x ) = 0 所 有 其 它 x 值 X \in \{x_1,\ x_2,\ \cdots \} \Rightarrow \begin{cases} p(x_i) > 0 & i = 1,\ 2,\ \cdots \\ p(x) = 0 & 所有其它x值 \end{cases} X∈{x1, x2, ⋯}⇒{p(xi)>0p(x)=0i=1, 2, ⋯所有其它x值
X ∈ { x 1 , x 2 , ⋯ } ⇒ ∑ i = 1 ∞ p ( x i ) = 1 X \in \{x_1,\ x_2,\ \cdots \} \Rightarrow \sum \limits^{\infty}_{i = 1} p(x_i) = 1 X∈{x1, x2, ⋯}⇒i=1∑∞p(xi)=1,故累积分布函数F可以用p(a)表示为 F ( a ) = ∑ ∀ x i ≤ a p ( x i ) F(a) = \sum \limits_{\forall x_i \leq a} p(x_i) F(a)=∀xi≤a∑p(xi)
例如:
假定X具有由
p
(
1
)
=
1
2
,
p
(
2
)
=
1
3
,
p
(
3
)
=
1
6
\\p(1) = \frac{1}{2},p(2) = \frac{1}{3},p(3) = \frac{1}{6}\newline
p(1)=21,p(2)=31,p(3)=61 给出的概率质量函数,则X的累积分布函数为
F
(
a
)
=
{
0
a
<
1
1
2
1
≤
a
<
2
1
3
2
≤
a
<
3
1
6
3
≤
a
F(a) = \begin{cases} 0 & a < 1 \\ \frac{1}{2} & 1 \leq a < 2 \\ \frac{1}{3} & 2 \leq a < 3 \\ \frac{1}{6} & 3 \leq a \end{cases}
F(a)=⎩⎪⎪⎪⎨⎪⎪⎪⎧0213161a<11≤a<22≤a<33≤a
2.2 伯努利随机变量:
试验的结果只有成功与失败,在成功时令X=1,在失败时令X=0,则X的概率质量函数
p
(
0
)
=
P
{
X
=
0
}
=
1
−
p
p
(
1
)
=
P
{
X
=
1
}
=
p
p(0) = P\{X = 0\} = 1 - p \\ p(1) = P\{X = 1\} = p\newline
p(0)=P{X=0}=1−pp(1)=P{X=1}=p随机变量X称为伯努利随机变量。
期望:
E
[
X
]
=
0
×
(
1
−
p
)
+
1
×
p
=
p
E[X] = 0 \times (1 - p) + 1 \times p = p
E[X]=0×(1−p)+1×p=p
方差:
D
(
X
)
=
E
(
X
2
)
−
[
E
(
X
)
]
2
=
p
−
p
2
=
p
(
1
−
p
)
D(X) = E(X^2) - [E(X)]^2 = p - p^2 = p(1-p)
D(X)=E(X2)−[E(X)]2=p−p2=p(1−p)
2.3 二项随机变量
假定做了n次独立试验,其中每次结果为成功的概率为p,失败的概率为
1
−
p
1 - p
1−p,如果以X代表出现在n次实验中成功的次数,那么X称为具有参数
(
n
,
p
)
(n, p)
(n,p)的二项随机变量。X的概率质量函数为
p
(
i
)
=
(
n
i
)
p
i
(
1
−
p
)
n
−
i
,
i
=
0
,
1
,
⋯
,
n
(
n
i
)
=
C
n
i
=
n
!
(
n
−
i
)
!
i
!
p(i) = \dbinom{n}{i} p^i (1 - p)^{n - i},\quad i = 0,\ 1,\ \cdots,\ n \\ \dbinom{n}{i} = C^{i}_{n} = \frac{n!}{(n - i)! i!}
p(i)=(in)pi(1−p)n−i,i=0, 1, ⋯, n(in)=Cni=(n−i)!i!n!
期望:
E
(
X
)
=
n
p
E(X) = np
E(X)=np
求解过程:
E
[
X
]
=
∑
i
=
0
n
i
p
(
i
)
=
∑
i
=
0
n
i
(
n
i
)
p
i
(
1
−
p
)
n
−
i
=
∑
i
=
1
n
i
n
!
(
n
−
i
)
!
i
!
p
i
(
1
−
p
)
n
−
i
=
∑
i
=
1
n
n
!
(
n
−
i
)
!
(
i
−
1
)
!
p
i
(
1
−
p
)
n
−
i
=
n
p
∑
i
=
1
n
(
n
−
1
)
!
(
n
−
i
)
!
(
i
−
1
)
!
p
i
−
1
(
1
−
p
)
n
−
i
=
n
p
∑
k
=
0
n
−
1
(
n
−
1
k
)
p
k
(
1
−
p
)
n
−
1
−
k
=
n
p
[
p
+
(
1
−
p
)
]
n
−
1
=
n
p
\begin{aligned} E[X] & = \sum^n_{i = 0} i p(i) \\ & = \sum^n_{i = 0} i \dbinom{n}{i} p^i (1 -p)^{n - i} \\ & = \sum^n_{i = 1} \frac{i n!}{(n - i)! i!} p^i (1 - p)^{n - i} \\ & = \sum^n_{i = 1} \frac{n!}{(n - i)! (i - 1)!} p^i (1 - p)^{n - i} \\ & = np \sum^n_{i = 1} \frac{(n - 1)!}{(n - i)! (i - 1)!} p^{i - 1}(1 - p)^{n - i} \\ & = np \sum^{n - 1}_{k = 0} \dbinom{n - 1}{k} p^k (1 - p)^{n - 1 - k} \\ & = np[p + (1 - p)]^{n - 1} \\ & = np \end{aligned}
E[X]=i=0∑nip(i)=i=0∑ni(in)pi(1−p)n−i=i=1∑n(n−i)!i!in!pi(1−p)n−i=i=1∑n(n−i)!(i−1)!n!pi(1−p)n−i=npi=1∑n(n−i)!(i−1)!(n−1)!pi−1(1−p)n−i=npk=0∑n−1(kn−1)pk(1−p)n−1−k=np[p+(1−p)]n−1=np
方差:
D
(
X
)
=
n
p
(
1
−
p
)
D(X) = np(1 - p)
D(X)=np(1−p)
求解过程:
E
[
X
2
]
=
∑
i
=
0
n
i
2
p
(
i
)
=
∑
i
=
0
n
i
2
(
n
i
)
p
i
(
1
−
p
)
n
−
i
=
n
p
∑
i
=
1
n
i
(
n
−
1
)
!
(
n
−
i
)
!
(
i
−
1
)
!
p
i
−
1
(
1
−
p
)
n
−
i
=
n
p
∑
k
=
0
n
−
1
(
k
+
1
)
(
n
−
1
)
!
(
n
−
1
−
k
)
!
k
!
p
k
(
1
−
p
)
n
−
1
−
k
=
n
p
[
∑
k
=
0
n
−
1
(
n
−
1
k
)
p
k
(
1
−
p
)
n
−
1
−
k
+
∑
k
=
0
n
−
1
k
(
n
−
1
)
!
(
n
−
1
−
k
)
!
k
!
p
k
(
1
−
p
)
n
−
1
−
k
]
=
n
p
[
1
+
(
n
−
1
)
p
]
D
(
X
)
=
E
(
X
2
)
−
E
2
(
X
)
=
n
p
[
1
+
(
n
−
1
)
p
]
−
(
n
p
)
2
=
n
p
(
1
−
p
)
\begin{aligned} E[X^2] & = \sum^n_{i = 0} i^2 p(i) \\ & = \sum^n_{i = 0} i^2 \dbinom{n}{i} p^i (1 -p)^{n - i} \\ & = np \sum^n_{i = 1} \frac{i (n - 1)!}{(n - i)! (i - 1)!} p^{i - 1}(1 - p)^{n - i} \\ & = np \sum^{n - 1}_{k = 0} \frac{(k + 1) (n - 1)!}{(n - 1 - k)! k!} p^k (1 - p)^{n - 1 - k} \\ & = np \left[ \sum^{n - 1}_{k = 0} \dbinom{n - 1}{k} p^k (1 - p)^{n - 1 - k} + \sum^{n - 1}_{k = 0} \frac{k (n - 1)!}{(n - 1 - k)! k!} p^k (1 - p)^{n - 1 - k} \right] \\ & = np[1 + (n - 1)p] \\ \end{aligned} \\ D(X) = E(X^2) - E^ 2 (X) = np[1 + (n - 1)p] - (np)^2 = np(1 - p)
E[X2]=i=0∑ni2p(i)=i=0∑ni2(in)pi(1−p)n−i=npi=1∑n(n−i)!(i−1)!i(n−1)!pi−1(1−p)n−i=npk=0∑n−1(n−1−k)!k!(k+1)(n−1)!pk(1−p)n−1−k=np[k=0∑n−1(kn−1)pk(1−p)n−1−k+k=0∑n−1(n−1−k)!k!k(n−1)!pk(1−p)n−1−k]=np[1+(n−1)p]D(X)=E(X2)−E2(X)=np[1+(n−1)p]−(np)2=np(1−p)
2.4 几何随机变量
假定进行独立试验直到出现一个结果为成功,其中每一个试验成功的概率都是p,如果以X记直到出现首次成功所需要做的试验次数,则称X为具有参数p的几何随机变量。概率质量函数为 p ( n ) = P { X = n } = ( 1 − p ) n − 1 p , n = 1 , 2 , ⋯ \\p(n) = P\{X = n\} = (1 - p)^{n - 1}p, \quad n =1,\ 2,\ \cdots p(n)=P{X=n}=(1−p)n−1p,n=1, 2, ⋯
期望:
E
(
X
)
=
1
p
E(X) = \frac{1}{p}
E(X)=p1
求解过程:
E
[
X
]
=
∑
n
=
1
∞
n
p
(
1
−
p
)
n
−
1
=
p
∑
n
=
1
∞
n
q
n
−
1
(
q
=
1
−
p
)
=
p
∑
n
=
1
∞
d
q
n
d
q
=
p
d
(
∑
n
=
1
∞
q
n
)
d
q
=
p
d
(
q
1
−
q
)
d
q
=
p
(
1
−
q
)
2
=
1
p
\begin{aligned} E[X] & = \sum^{\infty}_{n = 1}n p (1-p)^{n - 1} \\ & = p\sum^{\infty}_{n = 1} nq^{n - 1} (q = 1 - p) \\ & = p \sum^{\infty}_{n = 1} \frac{d q^n}{dq} \\ & = p \frac{d (\sum^{\infty}_{n = 1}q^n)}{dq} \\ & = p \frac{d(\frac{q}{1- q})}{dq} \\ & = \frac{p}{(1 - q)^2} \\ & = \frac{1}{p} \end{aligned}
E[X]=n=1∑∞np(1−p)n−1=pn=1∑∞nqn−1(q=1−p)=pn=1∑∞dqdqn=pdqd(∑n=1∞qn)=pdqd(1−qq)=(1−q)2p=p1
方差:
D
(
X
)
=
1
−
p
p
2
D(X) = \frac{1 - p}{p^2}
D(X)=p21−p
求解过程:
E
(
X
2
)
=
∑
n
=
1
∞
n
2
p
(
1
−
p
)
n
−
1
=
p
∑
n
=
1
∞
n
2
q
n
−
1
(
q
=
1
−
p
)
=
p
(
∑
n
=
1
∞
n
q
n
)
q
′
=
p
(
q
∑
n
=
1
∞
n
q
n
−
1
)
q
′
=
p
(
q
(
1
−
q
)
2
)
q
′
=
2
−
p
p
2
D
(
X
)
=
E
(
X
2
)
−
E
2
(
X
)
=
2
−
p
p
2
−
1
p
2
=
1
−
p
p
2
\begin{aligned} E(X^2) & = \sum^{\infty}_{n = 1} n^2 p (1-p)^{n - 1} \\ & = p\sum^{\infty}_{n = 1} n^2 q^{n - 1} (q = 1 - p) \\ & = p \left(\sum^{\infty}_{n = 1} n q^n \right)^{'}_q \\ & = p \left(q \sum^{\infty}_{n = 1} n q^{n - 1} \right)^{'}_q \\ & = p \left( \frac{q}{(1 - q)^2} \right)^{'}_q \\ & = \frac{2 - p}{p^2} \end{aligned} \\ D(X) = E(X^2) - E^2(X) = \frac{2 - p}{p^2} - \frac{1}{p^2} = \frac{1 - p}{p^2}
E(X2)=n=1∑∞n2p(1−p)n−1=pn=1∑∞n2qn−1(q=1−p)=p(n=1∑∞nqn)q′=p(qn=1∑∞nqn−1)q′=p((1−q)2q)q′=p22−pD(X)=E(X2)−E2(X)=p22−p−p21=p21−p
2.5 超几何随机变量
超几何分布: 从有限N个物件(包含M个指定种类的物件)中抽出n个物件,成功抽出该指定种类的物件的次数(不放回)
概率密度函数:
p
(
k
)
=
P
{
X
=
k
}
=
C
M
k
C
N
−
M
n
−
k
C
N
n
p(k) = P\{X = k\} = \frac{C^k_M C^{n - k}_{N - M}}{C^n_N}
p(k)=P{X=k}=CNnCMkCN−Mn−k
期望:
E
(
X
)
=
n
M
N
E(X) = \frac{nM}{N}
E(X)=NnM
求解过程:
方差:
D
(
X
)
=
n
M
N
(
1
−
M
N
)
(
1
−
n
−
1
N
−
1
)
D(X) = n \frac{M}{N} (1 - \frac{M}{N})(1 - \frac{n - 1}{N - 1})
D(X)=nNM(1−NM)(1−N−1n−1)
求解过程:
略
2.6 泊松随机变量
泊松随机变量:
∀
X
∈
N
,
∃
λ
>
0
,
p
(
i
)
=
P
{
X
=
i
}
=
e
−
λ
λ
i
i
!
i
=
0
,
1
,
⋯
\forall X \in N, \quad \exist \lambda > 0, \quad p(i) = P\{X = i\} = e^{-\lambda} \frac{\lambda^i}{i!} \quad i = 0,\ 1,\ \cdots
∀X∈N,∃λ>0,p(i)=P{X=i}=e−λi!λii=0, 1, ⋯
期望:
E
(
X
)
=
λ
E(X) = \lambda
E(X)=λ
求解过程:
E
[
X
]
=
∑
i
=
0
∞
i
e
−
λ
λ
i
i
!
=
∑
i
=
1
∞
e
−
λ
λ
i
(
i
−
1
)
!
=
λ
e
−
λ
∑
k
=
0
∞
λ
k
k
!
=
λ
e
−
λ
e
λ
=
λ
\begin{aligned} E[X] & = \sum^{\infty}_{i = 0} \frac{i e^{- \lambda} \lambda^i}{i !} \\ & = \sum^{\infty}_{i = 1} \frac{e^{- \lambda} \lambda^i}{(i - 1)!} \\ & = \lambda e^{- \lambda} \sum^{\infty}_{k = 0} \frac{\lambda^k}{k!} \\ & = \lambda e^{- \lambda} e^{\lambda} = \lambda \end{aligned}
E[X]=i=0∑∞i!ie−λλi=i=1∑∞(i−1)!e−λλi=λe−λk=0∑∞k!λk=λe−λeλ=λ
方差:
D
(
X
)
=
λ
D(X) = \lambda
D(X)=λ
求解过程:
E
(
X
2
)
=
∑
i
=
0
∞
i
2
e
−
λ
λ
i
i
!
=
λ
∑
i
=
1
∞
i
e
−
λ
λ
i
−
1
(
i
−
1
)
!
=
λ
(
e
−
λ
∑
k
=
0
∞
λ
k
k
!
+
λ
∫
k
=
1
∞
e
−
λ
λ
k
−
1
(
k
−
1
)
!
)
=
λ
(
e
−
λ
e
λ
+
λ
e
−
λ
e
λ
)
=
λ
(
1
+
λ
)
D
(
X
)
=
E
(
X
2
)
−
E
2
(
X
)
=
λ
(
1
+
λ
)
−
λ
2
=
λ
\begin{aligned} E(X^2) & = \sum^{\infty}_{i = 0} \frac{i^2 e^{- \lambda} \lambda^i}{i !} \\ & = \lambda \sum^{\infty}_{i = 1} \frac{i e^{- \lambda} \lambda^{i - 1}}{(i - 1)!} \\ & = \lambda \left(e^{- \lambda} \sum^{\infty}_{k = 0} \frac{\lambda^k}{k!} + \lambda \int \limits^{\infty}_{k = 1} \frac{e^{-\lambda} \lambda ^{k - 1}}{(k - 1)!} \right) \\ & = \lambda \left(e^{- \lambda} e^{\lambda} + \lambda e^{-\lambda} e^\lambda \right) \\ & = \lambda (1 + \lambda) \end{aligned} \\ D(X) = E(X^2) - E^2(X) = \lambda(1 + \lambda) - \lambda^2 = \lambda
E(X2)=i=0∑∞i!i2e−λλi=λi=1∑∞(i−1)!ie−λλi−1=λ⎝⎛e−λk=0∑∞k!λk+λk=1∫∞(k−1)!e−λλk−1⎠⎞=λ(e−λeλ+λe−λeλ)=λ(1+λ)D(X)=E(X2)−E2(X)=λ(1+λ)−λ2=λ
例如:
2.7 总结
离散概率分布 | 概率质量函数 p ( x ) p(x) p(x) | 矩母函数 ϕ ( t ) \phi(t) ϕ(t) | 期望 E ( X ) E(X) E(X) | 方差 D ( X ) D(X) D(X) |
---|---|---|---|---|
离散随机变量 | p ( a ) = P { X = a } p(a) = P\{X = a\} p(a)=P{X=a} | ∑ x e t x p ( x ) \sum_x e^{tx} p(x) x∑etxp(x) | ∑ x x p ( x ) ϕ ′ ( 0 ) \sum_x x p(x) \\ \phi^\prime(0) x∑xp(x)ϕ′(0) | E ( X 2 ) − E 2 ( X ) ϕ ′ ′ ( 0 ) − [ ϕ ′ ( 0 ) ] 2 E(X^2) - E^2(X) \\ \phi^{\prime \prime}(0) - [\phi^\prime(0)]^2 E(X2)−E2(X)ϕ′′(0)−[ϕ′(0)]2 |
伯努利分布 | p ( 0 ) = 1 − p p ( 1 ) = p p(0) = 1 - p \\ p(1) = p p(0)=1−pp(1)=p | p e t + ( 1 − p ) pe^t + (1 - p) pet+(1−p) | p p p | p ( 1 − p ) p(1 - p) p(1−p) |
二项分布 | p ( x ) = ( n x ) p x ( 1 − p ) n − x p(x) = \dbinom{n}{x} p^x (1 - p)^{n - x} p(x)=(xn)px(1−p)n−x | ( p e t + ( 1 − p ) ) n (pe^t + (1 - p))^n (pet+(1−p))n | n p np np | n p ( 1 − p ) np(1 - p) np(1−p) |
几何分布 | p ( x ) = ( 1 − p ) x − 1 p p(x) = (1 - p)^{x - 1}p p(x)=(1−p)x−1p | p e t 1 − ( 1 − p ) e t \frac{p e^t}{1 - (1 - p) e^t} 1−(1−p)etpet | 1 p \frac{1}{p} p1 | 1 − p p 2 \frac{1 - p}{p^2} p21−p |
超几何分布 | p ( x ) = C M x C N − M n − x C N n p(x) = \frac{C^x_M C^{n - x}_{N - M}}{C^n_N} p(x)=CNnCMxCN−Mn−x | ∑ x e t x p ( x ) \sum_x e^{tx} p(x) x∑etxp(x) | n M N \frac{nM}{N} NnM | n M N ( 1 − M N ) ( 1 − n − 1 N − 1 ) n\frac{M}{N}(1 - \frac{M}{N})(1 - \frac{n - 1}{N - 1}) nNM(1−NM)(1−N−1n−1) |
泊松分布 | p ( x ) = e − λ λ x x ! p(x) = e^{-\lambda} \frac{\lambda^x}{x!} p(x)=e−λx!λx | e λ ( e t − 1 ) e^{\lambda(e^t - 1)} eλ(et−1) | λ \lambda λ | λ \lambda λ |
3. 连续随机变量、期望、方差
3.1 连续随机变量
一个随机变量的可能值是不可数的,记为X,则X是一个连续的随机变量。 ∃ f ( x ) ≥ 0 且 x ∈ ( − ∞ , ∞ ) , 使 ∀ 实 数 X ∈ B \exist f(x) \geq 0且x \in (-\infty,\ \infty),使\forall 实数X \in B ∃f(x)≥0且x∈(−∞, ∞),使∀实数X∈B,有X的概率密度函数为 P { X ∈ B } = ∫ B f ( x ) d x P\{X \in B\} = \int_Bf(x) dx P{X∈B}=∫Bf(x)dx
f ( x ) f(x) f(x)必定满足 1 = P { X ∈ ( − ∞ , ∞ ) } = ∫ − ∞ ∞ f ( x ) d x 1 = P\{X \in (-\infty,\ \infty)\} = \int^{\infty}_{-\infty} f(x) dx 1=P{X∈(−∞, ∞)}=∫−∞∞f(x)dx
例如:设 B = [ a , b ] , 则 P { a ≤ X ≤ b } = ∫ a b f ( x ) d x B = [a,\ b],\ 则P\{a \leq X \leq b\} = \int^{b}_{a} f(x) dx B=[a, b], 则P{a≤X≤b}=∫abf(x)dx
累积分布函数F和 f f f的关系表示为 F ( a ) = P { X ∈ ( − ∞ , a ] } = ∫ − ∞ a f ( x ) d x \\F(a) = P\{X \in (-\infty, a]\} = \int \limits^a_{-\infty} f(x) dx F(a)=P{X∈(−∞,a]}=−∞∫af(x)dx
密度函数是累积分布函数的导数
期望:
E
[
x
]
=
∫
−
∞
∞
x
f
(
x
)
d
x
E[x] = \int \limits^{\infty}_{- \infty} xf(x) dx
E[x]=−∞∫∞xf(x)dx
方差:
D
(
X
)
=
E
(
X
2
)
−
E
2
(
X
)
=
∫
−
∞
∞
x
2
f
(
x
)
d
x
−
[
∫
−
∞
∞
x
f
(
x
)
d
x
]
2
D(X) = E(X^2) - E^2(X) = \int \limits^{\infty}_{- \infty} x^2 f(x) dx - \left[ \int \limits^{\infty}_{- \infty} x f(x) dx \right]^2
D(X)=E(X2)−E2(X)=−∞∫∞x2f(x)dx−[−∞∫∞xf(x)dx]2
3.2 均匀随机变量
均匀随机变量: 一个随机变量X在 ( α , β ) (\alpha,\ \beta) (α, β)的任意特定子区间的概率等于该子区间的长度。
X是区间
(
α
,
β
)
(\alpha,\ \beta)
(α, β)上的均匀随机变量,则概率密度函数为
f
(
x
)
=
{
1
β
−
α
α
<
x
<
β
0
其
它
f(x) = \begin{cases} \frac{1}{\beta - \alpha} & \alpha < x < \beta \\ 0 & 其它 \end{cases}
f(x)={β−α10α<x<β其它
期望:
E
(
X
)
=
β
+
α
2
E(X) = \frac{\beta + \alpha}{2}
E(X)=2β+α
求解过程:
E
(
X
)
=
∫
α
β
x
β
−
α
d
x
=
x
2
2
(
β
−
α
)
∣
α
β
=
β
+
α
2
\begin{aligned} E(X) & = \int^\beta_\alpha \frac{x}{\beta - \alpha} dx \\ & = \frac{x^2}{2(\beta - \alpha)} |^\beta_\alpha \\ & = \frac{\beta + \alpha}{2} \end{aligned}
E(X)=∫αββ−αxdx=2(β−α)x2∣αβ=2β+α
方差:
D
(
X
)
=
(
β
−
α
)
2
12
D(X) = \frac{(\beta - \alpha)^2}{12}
D(X)=12(β−α)2
求解过程:
E
(
X
2
)
=
∫
α
β
x
2
β
−
α
d
x
=
x
3
3
(
β
−
α
)
∣
α
β
=
β
2
+
α
2
+
α
β
3
D
(
X
)
=
E
(
X
2
)
−
E
2
(
X
)
=
β
2
+
α
2
+
α
β
3
−
[
β
−
α
2
]
2
=
(
β
−
α
)
2
12
\begin{aligned} E(X^2) & = \int \limits^\beta_\alpha \frac{x^2}{\beta - \alpha} dx \\ & = \frac{x^3}{3(\beta - \alpha)} |^\beta_\alpha \\ & = \frac{\beta^2 + \alpha_2 + \alpha\beta}{3} \end{aligned} \\ D(X) = E(X^2) - E^2(X) = \frac{\beta^2 + \alpha^2 + \alpha\beta}{3} - [\frac{\beta - \alpha}{2}]^2 = \frac{(\beta - \alpha)^2}{12}
E(X2)=α∫ββ−αx2dx=3(β−α)x3∣αβ=3β2+α2+αβD(X)=E(X2)−E2(X)=3β2+α2+αβ−[2β−α]2=12(β−α)2
\newline
例如:
1. 计算均匀分布在
(
α
,
β
)
(\alpha, \beta)
(α,β)上的随机变量的累积分布函数。
解:
f
(
x
)
=
{
1
β
−
α
α
<
x
<
β
0
其
它
f(x) = \begin{cases} \frac{1}{\beta - \alpha} & \alpha < x < \beta \\ 0 & 其它 \end{cases}
f(x)={β−α10α<x<β其它
累积分布函数
F
(
a
)
=
∫
−
∞
a
f
(
x
)
d
x
F(a) = \int^a_{- \infty} f(x) dx
F(a)=∫−∞af(x)dx
当
a
≤
α
a \leq \alpha
a≤α时,
F
(
a
)
=
∫
−
∞
a
f
(
x
)
d
x
=
∫
−
∞
a
0
d
x
=
0
F(a) = \int^a_{- \infty} f(x) dx = \int^a_{- \infty} 0 dx = 0
F(a)=∫−∞af(x)dx=∫−∞a0dx=0
当
a
<
β
a < \beta
a<β时,
F
(
a
)
=
∫
−
∞
a
f
(
x
)
d
x
=
∫
−
∞
α
f
(
x
)
d
x
+
∫
α
a
f
(
x
)
d
x
=
0
+
1
β
−
α
x
∣
α
a
=
a
−
α
β
−
α
F(a) = \int^a_{- \infty} f(x) dx = \int^{\alpha}_{- \infty} f(x) dx + \int^{a}_{\alpha} f(x) dx = 0 + \frac{1}{\beta - \alpha} x |^a_\alpha = \frac{a - \alpha}{\beta - \alpha}
F(a)=∫−∞af(x)dx=∫−∞αf(x)dx+∫αaf(x)dx=0+β−α1x∣αa=β−αa−α
当
a
≥
β
a \geq \beta
a≥β时,
F
(
a
)
=
∫
−
∞
a
f
(
x
)
d
x
=
∫
−
∞
α
f
(
x
)
d
x
+
∫
α
β
f
(
x
)
d
x
+
∫
β
a
f
(
x
)
d
x
=
0
+
1
β
−
α
∣
α
β
+
0
=
1
F(a) = \int^{a}_{- \infty} f(x) dx = \int^{\alpha}_{- \infty} f(x) dx + \int^{\beta}_{ \alpha} f(x) dx + \int^{a}_{\beta} f(x) dx = 0 + \frac{1}{\beta - \alpha} |^\beta_\alpha + 0 = 1
F(a)=∫−∞af(x)dx=∫−∞αf(x)dx+∫αβf(x)dx+∫βaf(x)dx=0+β−α1∣αβ+0=1
综上有
F
(
a
)
=
{
0
a
≤
α
a
−
α
β
−
α
α
<
a
<
β
1
α
≥
β
F(a) = \begin{cases} 0 & a \leq \alpha \\ \frac{a - \alpha}{\beta - \alpha} & \alpha < a < \beta \\ 1 & \alpha \geq \beta \end{cases}
F(a)=⎩⎪⎨⎪⎧0β−αa−α1a≤αα<a<βα≥β
\newline
2. X均匀分布在
(
0
,
10
)
(0,\ 10)
(0, 10)上,计算概率
(
a
)
X
<
3
(
b
)
X
>
7
(
c
)
1
<
X
<
6
(a) X < 3 \quad (b) X > 7 \quad (c) 1 < X < 6
(a)X<3(b)X>7(c)1<X<6
解:
f
(
x
)
=
{
1
10
0
<
x
<
10
0
其
它
f(x) = \begin{cases} \frac{1}{10} & 0 < x < 10 \\ 0 & 其它 \end{cases}
f(x)={10100<x<10其它
则
P
{
X
<
3
}
=
∫
−
∞
3
f
(
x
)
d
x
=
∫
−
∞
0
0
d
x
+
∫
0
3
1
10
d
x
=
3
10
P
{
X
>
7
}
=
∫
−
∞
∞
f
(
x
)
d
x
−
∫
−
∞
7
f
(
x
)
d
x
=
∫
0
10
1
10
d
x
−
∫
0
7
1
10
d
x
=
3
10
P
{
1
<
X
<
6
}
=
∫
1
6
1
10
d
x
=
1
2
P\{X < 3\} = \int^3_{- \infty} f(x) dx = \int^0_{- \infty} 0 dx + \int^3_0 \frac{1}{10} dx = \frac{3}{10} \\ P\{X > 7\} = \int^{\infty}_{- \infty} f(x) dx - \int^{7}_{- \infty} f(x) dx = \int^{10}_0 \frac{1}{10} dx - \int^{7}_{0} \frac{1}{10} dx = \frac{3}{10} \\ P\{1 < X < 6\} = \int^6_1 \frac{1}{10} dx = \frac{1}{2}
P{X<3}=∫−∞3f(x)dx=∫−∞00dx+∫03101dx=103P{X>7}=∫−∞∞f(x)dx−∫−∞7f(x)dx=∫010101dx−∫07101dx=103P{1<X<6}=∫16101dx=21
3.3 指数随机变量
若一个连续随机变量的概率密度函数给定为,对于某个
λ
>
0
\lambda > 0
λ>0
f
(
x
)
=
{
λ
e
−
λ
x
x
≥
0
0
x
<
0
f(x) = \begin{cases} \lambda e^ {- \lambda x} & x \geq 0 \\ 0 & x < 0 \end{cases}
f(x)={λe−λx0x≥0x<0
则称其为具有参数
λ
\lambda
λ的指数随机变量。
累积分布函数F为
F
(
a
)
=
∫
0
a
λ
e
−
λ
x
d
x
=
1
−
e
−
λ
a
a
≥
0
F(a) = \int \limits^a_0 \lambda e^{- \lambda x} dx = 1 - e^{- \lambda a} \quad a \geq 0
F(a)=0∫aλe−λxdx=1−e−λaa≥0
期望:
E
(
X
)
=
1
λ
E(X) = \frac{1}{\lambda}
E(X)=λ1
求解过程:
E
(
X
)
=
∫
0
∞
x
λ
e
−
λ
x
d
x
=
∫
0
∞
−
x
d
(
e
−
λ
x
)
=
−
x
e
−
λ
x
∣
0
∞
+
∫
0
∞
e
−
λ
x
d
x
=
−
x
e
−
λ
x
∣
0
∞
−
1
λ
e
−
λ
x
∣
0
∞
=
1
λ
\begin{aligned} E(X) & = \int^\infty_0 x \lambda e^{- \lambda x} dx \\ & = \int^\infty_0 -x d(e^{- \lambda x}) \\ & = -x e^{- \lambda x} |^\infty_0 + \int^\infty_0 e^{- \lambda x} dx \\ & = -x e^{- \lambda x} |^\infty_0 - \frac{1}{\lambda} e^{- \lambda x} |^\infty_0 \\ & = \frac{1}{\lambda} \end{aligned}
E(X)=∫0∞xλe−λxdx=∫0∞−xd(e−λx)=−xe−λx∣0∞+∫0∞e−λxdx=−xe−λx∣0∞−λ1e−λx∣0∞=λ1
方差:
E
(
X
2
)
=
∫
0
∞
x
2
λ
e
−
λ
x
d
x
=
∫
0
∞
−
x
2
d
(
e
−
λ
x
)
=
−
x
2
e
−
λ
x
∣
0
∞
+
∫
0
∞
e
−
λ
x
d
x
2
=
−
x
2
e
−
λ
x
∣
0
∞
−
2
λ
[
x
e
−
λ
x
∣
0
∞
+
1
λ
e
−
λ
x
∣
0
∞
]
=
0
−
2
λ
(
0
−
1
λ
)
=
2
λ
2
D
(
X
)
=
E
(
X
2
)
−
E
2
(
X
)
=
2
λ
2
−
(
1
λ
)
2
=
1
λ
2
\begin{aligned} E(X^2) & = \int^\infty_0 x^2 \lambda e^{- \lambda x} dx \\ & = \int^\infty_0 -x^2 d(e^{- \lambda x}) \\ & = -x^2 e^{- \lambda x} |^\infty_0 + \int^\infty_0 e^{- \lambda x} dx^2 \\ & = -x^2 e^{- \lambda x} |^\infty_0 - \frac{2}{\lambda} \left[x e^{- \lambda x} |^\infty_0 + \frac{1}{\lambda} e^{- \lambda x} |^\infty_0 \right] \\ & = 0 - \frac{2}{\lambda} (0 - \frac{1}{\lambda}) \\ & = \frac{2}{\lambda^2} \end{aligned} \\ D(X) = E(X^2) - E^2(X) = \frac{2}{\lambda^2} - (\frac{1}{\lambda})^2 = \frac{1}{\lambda^2}
E(X2)=∫0∞x2λe−λxdx=∫0∞−x2d(e−λx)=−x2e−λx∣0∞+∫0∞e−λxdx2=−x2e−λx∣0∞−λ2[xe−λx∣0∞+λ1e−λx∣0∞]=0−λ2(0−λ1)=λ22D(X)=E(X2)−E2(X)=λ22−(λ1)2=λ21
3.4 伽马随机变量
密度函数给定为,对于
λ
>
0
,
α
>
0
\lambda > 0,\alpha > 0
λ>0,α>0,
f
(
x
)
=
{
λ
e
−
λ
x
(
λ
x
)
α
−
1
Γ
(
α
)
x
≥
0
0
x
<
0
f(x) = \begin{cases} \frac{\lambda e^{- \lambda x}(\lambda x) ^ {\alpha - 1}}{\Gamma(\alpha)} & x \geq 0 \\ 0 & x < 0 \end{cases}
f(x)={Γ(α)λe−λx(λx)α−10x≥0x<0的连续随机变量,称为具有参数
λ
\lambda
λ和
α
\alpha
α的伽马随机变量。
Γ
(
α
)
\Gamma(\alpha)
Γ(α)称为伽马函数,定义为
Γ
(
α
)
=
∫
0
∞
e
−
x
x
α
−
1
d
x
\Gamma(\alpha) = \int^\infty_0 e^{-x} x^{\alpha - 1} dx
Γ(α)=∫0∞e−xxα−1dx
3.5 正态随机变量
X是具有参数
μ
\mu
μ和
σ
2
\sigma^2
σ2的正态随机变量(或者简单地说,X是正态地分布),X的概率密度函数为
f
(
x
)
=
1
2
π
σ
e
−
(
x
−
μ
)
2
2
σ
2
f(x) = \frac{1}{\sqrt{2 \pi} \sigma} e^{\frac{- (x - \mu)^2}{2 \sigma^2}}
f(x)=2πσ1e2σ2−(x−μ)2
如果X以参数 μ \mu μ和 σ 2 \sigma^2 σ2正态地分布,那么 Y = α X + β Y = \alpha X + \beta Y=αX+β以参数 α μ + β \alpha \mu + \beta αμ+β和 α 2 σ 2 \alpha^2 \sigma^2 α2σ2正态地分布
期望:
E
(
X
)
=
u
E(X) = u
E(X)=u
求解过程:
E
(
X
)
=
1
2
π
σ
∫
−
∞
∞
x
e
−
(
x
−
μ
)
2
2
σ
2
d
x
=
1
2
π
σ
∫
−
∞
∞
(
x
−
μ
)
e
−
(
x
−
u
)
2
2
σ
2
d
x
+
μ
1
2
π
σ
∫
−
∞
∞
e
−
(
x
−
μ
)
2
2
σ
2
d
x
=
1
2
π
σ
∫
−
∞
∞
y
e
−
y
2
2
σ
2
d
y
+
μ
∫
−
∞
∞
f
(
x
)
d
x
=
μ
∫
−
∞
∞
f
(
x
)
d
x
=
μ
\begin{aligned} E(X) & = \frac{1}{\sqrt{2 \pi} \sigma} \int^\infty_{- \infty} x e^{\frac{-(x - \mu)^2}{2 \sigma^2}} dx \\ & = \frac{1}{\sqrt{2 \pi} \sigma} \int^\infty_{- \infty} (x - \mu) e^{\frac{-(x - u)^2}{2 \sigma^2}} dx + \mu \frac{1}{\sqrt{2 \pi} \sigma} \int^\infty_{- \infty} e^{\frac{-(x - \mu)^2}{2 \sigma^2}} dx \\ & = \frac{1}{\sqrt{2 \pi} \sigma} \int^\infty_{- \infty} y e^{\frac{-y^2}{2 \sigma^2}} dy + \mu \int^\infty_{- \infty} f(x) dx \\ & = \mu \int^\infty_{- \infty} f(x) dx \\ & = \mu \end{aligned}
E(X)=2πσ1∫−∞∞xe2σ2−(x−μ)2dx=2πσ1∫−∞∞(x−μ)e2σ2−(x−u)2dx+μ2πσ1∫−∞∞e2σ2−(x−μ)2dx=2πσ1∫−∞∞ye2σ2−y2dy+μ∫−∞∞f(x)dx=μ∫−∞∞f(x)dx=μ
方差:
D
(
X
)
=
σ
2
D(X) = \sigma^2
D(X)=σ2
求解过程:
略
3.6 总结
连续概率分布 | 概率密度函数 f ( x ) f(x) f(x) | 矩母函数 ϕ ( t ) \phi(t) ϕ(t) | 期望 E ( X ) E(X) E(X) | 方差 D ( X ) D(X) D(X) |
---|---|---|---|---|
连续随机变量 | P { X ∈ B } = ∫ B f ( x ) d x P\{X \in B\} = \int_Bf(x) dx P{X∈B}=∫Bf(x)dx | ∫ B e t x f ( x ) d x \int_B e^{tx} f(x) dx ∫Betxf(x)dx | ∫ − ∞ ∞ x f ( x ) d x ϕ ′ ( 0 ) \int_{-\infty}^{\infty}x f(x) dx\\ \phi^\prime(0) ∫−∞∞xf(x)dxϕ′(0) | E ( X 2 ) − E 2 ( X ) ϕ ′ ′ ( 0 ) − [ ϕ ′ ( 0 ) ] 2 E(X^2) - E^2(X) \\ \phi^{\prime \prime}(0) - [\phi^\prime(0)]^2 E(X2)−E2(X)ϕ′′(0)−[ϕ′(0)]2 |
均匀分布 | f ( x ) = { 1 b − a a < x < b 0 其 它 f(x) = \begin{cases} \frac{1}{b - a} & a < x < b \\0 & 其它 \end{cases} f(x)={b−a10a<x<b其它 | e b t − e a t ( b − a ) t \frac{e^{bt} - e^{at}}{(b - a)t} (b−a)tebt−eat | a + b 2 \frac{a + b}{2} 2a+b | ( b − a ) 2 12 \frac{(b - a)^2}{12} 12(b−a)2 |
指数分布 | f ( x ) = { λ e − λ x x ≥ 0 0 x < 0 f(x) = \begin{cases} \lambda e^ {- \lambda x} & x \geq 0 \\ 0 & x < 0 \end{cases} f(x)={λe−λx0x≥0x<0 | λ λ − t \frac{\lambda}{\lambda - t} λ−tλ | 1 λ \frac{1}{\lambda} λ1 | 1 λ 2 \frac{1}{\lambda^2} λ21 |
伽马分布 | f ( x ) = { λ e − λ x ( λ x ) α − 1 Γ ( α ) x ≥ 0 0 x < 0 f(x) = \begin{cases} \frac{\lambda e^{- \lambda x}(\lambda x) ^ {\alpha - 1}}{\Gamma(\alpha)} & x \geq 0 \\ 0 & x < 0 \end{cases} f(x)={Γ(α)λe−λx(λx)α−10x≥0x<0 | ( λ λ − t ) n \begin{pmatrix}\frac{\lambda}{\lambda - t}\end{pmatrix}^n (λ−tλ)n | n λ \frac{n}{\lambda} λn | n λ 2 \frac{n}{\lambda^2} λ2n |
正态分布 | f ( x ) = 1 2 π σ e − ( x − μ ) 2 2 σ 2 f(x) = \frac{1}{\sqrt{2 \pi} \sigma} e^{\frac{- (x - \mu)^2}{2 \sigma^2}} f(x)=2πσ1e2σ2−(x−μ)2 | e μ t + σ 2 t 2 2 e^{\mu t + \frac{\sigma^2 t^2}{2}} eμt+2σ2t2 | μ \mu μ | σ 2 \sigma^2 σ2 |
4. 联合分布的随机变量
任意两个随机变量X和Y,定义X和Y的联合累积概率分布函数为
F
(
a
,
b
)
=
p
{
X
≤
a
,
Y
≤
b
}
−
∞
<
a
,
b
<
∞
F(a,\ b) = p\{X \leq a,\ Y \leq b\} \qquad -\infty < a, b < \infty
F(a, b)=p{X≤a, Y≤b}−∞<a,b<∞
X的分布可以由X和Y的联合分布得到
F
X
(
a
)
=
P
{
X
≤
a
}
=
P
{
X
≤
a
,
Y
<
∞
}
=
F
(
a
,
∞
)
F_X(a) = P\{X \leq a\} = P\{X \leq a,\ Y < \infty\} = F(a,\ \infty)
FX(a)=P{X≤a}=P{X≤a, Y<∞}=F(a, ∞)
Y的分布可以由X和Y的联合分布得到
F
Y
(
b
)
=
P
{
Y
≤
b
}
=
P
{
X
<
∞
,
Y
≤
b
}
=
F
(
∞
,
b
)
F_Y(b) = P\{Y \leq b\} = P\{X < \infty,\ Y \leq b\} = F(\infty,\ b)
FY(b)=P{Y≤b}=P{X<∞, Y≤b}=F(∞, b)
联合概率密度函数:
∃
f
(
x
,
y
)
,
∀
x
,
y
∈
R
,
有
P
{
X
∈
A
,
Y
∈
B
}
=
∫
B
∫
A
f
(
x
,
y
)
d
x
d
y
\exist f(x, y),\ \forall x,\ y \in R,\ 有P\{X \in A,\ Y \in B\} = \int \limits_B \int \limits_A f(x,\ y) dx dy
∃f(x,y), ∀x, y∈R, 有P{X∈A, Y∈B}=B∫A∫f(x, y)dxdy
X的概率密度函数:
f
X
(
x
)
=
∫
−
∞
∞
f
(
x
,
y
)
d
y
f_X(x) = \int \limits^\infty_{- \infty} f(x,\ y) dy
fX(x)=−∞∫∞f(x, y)dy
Y的概率密度函数:
f
Y
(
y
)
=
∫
−
∞
∞
f
(
x
,
y
)
d
x
f_Y(y) = \int \limits^\infty_{- \infty} f(x,\ y) dx
fY(y)=−∞∫∞f(x, y)dx
函数的概率密度函数:
E
[
g
(
X
,
Y
)
]
=
{
∑
y
∑
x
g
(
x
,
y
)
p
(
x
,
y
)
离
散
情
形
∫
−
∞
∞
∫
−
∞
∞
g
(
x
,
y
)
f
(
x
,
y
)
d
x
d
y
连
续
情
形
E[g(X,\ Y)] = \begin{cases} \sum \limits_y \sum \limits_x g(x,\ y) p(x,\ y) & 离散情形 \\ \int \limits^\infty_{- \infty} \int \limits^\infty_{- \infty} g(x,\ y) f(x,\ y) dx dy & 连续情形 \end{cases} \newline
E[g(X, Y)]=⎩⎪⎨⎪⎧y∑x∑g(x, y)p(x, y)−∞∫∞−∞∫∞g(x, y)f(x, y)dxdy离散情形连续情形
E ( ∑ i = 1 n a i X i ) = ∑ i = 1 n a i E ( X i ) E(\sum \limits_{i = 1}^n a_iX_i) = \sum \limits^n_{i = 1}a_i E(X_i) E(i=1∑naiXi)=i=1∑naiE(Xi)
例如:
独立随机变量:
∀
a
,
b
,
有
P
{
X
≤
a
,
Y
≤
b
}
=
P
{
X
≤
a
}
P
{
Y
≤
b
}
\forall a,\ b,\ \ 有P\{X \leq a,\ Y \leq b\} = P\{X \leq a\} P\{Y \leq b\}
∀a, b, 有P{X≤a, Y≤b}=P{X≤a}P{Y≤b}
若X和Y为独立随机变量,其联合分布函数为F,则
F
(
a
,
b
)
=
F
X
(
a
)
F
Y
(
b
)
F(a,\ b) = F_X(a) F_Y(b)
F(a, b)=FX(a)FY(b)。
当X和Y都是离散时,独立的条件简化为
p
(
x
,
y
)
=
p
X
(
x
)
p
Y
(
y
)
p(x,\ y) = p_X(x) p_Y(y)
p(x, y)=pX(x)pY(y)
当X和Y都是连续时,独立的条件简化为
f
(
x
y
)
=
f
X
(
x
)
f
Y
(
y
)
f(x\ y) = f_X(x) f_Y(y)
f(x y)=fX(x)fY(y)
若果X和Y是独立的,那么对于任意函数g和h,有
E
[
g
(
X
)
h
(
Y
)
]
=
E
[
g
(
X
)
]
E
[
h
(
Y
)
]
E[g(X)h(Y)] = E[g(X)] E[h(Y)]
E[g(X)h(Y)]=E[g(X)]E[h(Y)]
协方差:
C
o
v
(
X
,
Y
)
=
E
[
(
X
−
E
[
X
]
)
(
Y
−
E
[
Y
]
)
]
=
E
[
X
Y
−
Y
E
[
X
]
−
X
E
[
Y
]
+
E
[
X
]
E
[
Y
]
]
=
E
[
X
Y
]
−
E
[
Y
]
E
[
X
]
−
E
[
X
]
E
[
Y
]
+
E
[
X
]
E
[
Y
]
=
E
[
X
Y
]
−
E
[
X
]
E
[
Y
]
\begin{aligned} Cov(X,\ Y) & = E[(X - E[X])(Y - E[Y])] \\ & = E[XY - YE[X] - XE[Y] + E[X]E[Y]] \\ & = E[XY] - E[Y]E[X] - E[X]E[Y] + E[X]E[Y] \\ & = E[XY] - E[X]E[Y] \end{aligned}
Cov(X, Y)=E[(X−E[X])(Y−E[Y])]=E[XY−YE[X]−XE[Y]+E[X]E[Y]]=E[XY]−E[Y]E[X]−E[X]E[Y]+E[X]E[Y]=E[XY]−E[X]E[Y]
若X和Y独立,则
C
o
v
(
X
,
Y
)
=
0
Cov(X,\ Y) = 0
Cov(X, Y)=0
C
o
v
(
X
,
Y
)
>
0
⇒
Cov(X,\ Y) > 0 \Rightarrow
Cov(X, Y)>0⇒ X增加时,Y倾向于增加
C
o
v
(
X
,
Y
)
<
0
⇒
\newline Cov(X,\ Y) < 0 \Rightarrow
Cov(X, Y)<0⇒ X增加时,Y倾向于减少
C o v ( X , X ) = V a r ( X ) C o v ( X , Y ) = C o v ( Y , X ) C o v ( c X , Y ) = c C o v ( X , Y ) C o v ( X , Y + Z ) = C o v ( X , Y ) + C o v ( X , Z ) C o v ( ∑ i = 1 n X i , ∑ i = 1 n Y i ) = ∑ i = 1 n ∑ i = 1 n C o v ( X i , Y i ) Cov(X,\ X) = Var(X) \newline Cov(X,\ Y) =Cov(Y,\ X) \newline Cov(cX,\ Y) = cCov(X,\ Y) \newline Cov(X,\ Y + Z) = Cov(X,\ Y) + Cov(X,\ Z) \newline Cov(\sum \limits^n_{i = 1}X_i,\ \sum \limits^n_{i = 1} Y_i) = \sum \limits^n_{i = 1} \sum \limits^n_{i = 1} Cov(X_i,\ Y_i)\newline Cov(X, X)=Var(X)Cov(X, Y)=Cov(Y, X)Cov(cX, Y)=cCov(X, Y)Cov(X, Y+Z)=Cov(X, Y)+Cov(X, Z)Cov(i=1∑nXi, i=1∑nYi)=i=1∑ni=1∑nCov(Xi, Yi)
V
a
r
(
∑
i
=
1
n
X
i
)
=
∑
i
=
1
n
V
a
r
(
X
i
)
+
2
∑
i
=
1
n
∑
j
<
i
C
o
v
(
X
i
,
X
j
)
Var(\sum \limits^n_{i = 1}X_i) = \sum \limits^n_{i = 1} Var(X_i) + 2\sum \limits^n_{i = 1}\sum \limits_{j < i}Cov(X_i, X_j)
Var(i=1∑nXi)=i=1∑nVar(Xi)+2i=1∑nj<i∑Cov(Xi,Xj)
当
X
i
X_i
Xi是独立随机变量时,
V
a
r
(
∑
i
=
1
n
X
i
)
=
∑
i
=
1
n
V
a
r
(
X
i
)
Var(\sum \limits^n_{i = 1} X_i) = \sum \limits^n_{i = 1}Var(X_i)
Var(i=1∑nXi)=i=1∑nVar(Xi)
这两个公式常用来计算方差
例如:
样本均值: 若
X
1
,
⋯
,
X
n
X_1,\ \cdots,\ X_n
X1, ⋯, Xn独立同分布,则随机变量
X
‾
=
∑
i
=
1
n
X
i
n
\overline X = \sum \limits^n_{i = 1} \frac{X_i}{n}
X=i=1∑nnXi就是样本均值
均值是总和/个数,期望是带上概率计算出来的“均值”
X
1
,
⋯
,
X
n
独
立
同
分
布
,
E
(
X
)
=
μ
,
D
(
X
)
=
σ
2
X_1,\ \cdots,\ X_n独立同分布,E(X) = \mu,D(X) = \sigma^2
X1, ⋯, Xn独立同分布,E(X)=μ,D(X)=σ2,则
1.
E
[
X
‾
]
=
μ
2.
V
a
r
(
X
‾
)
=
σ
2
n
3.
C
o
v
(
X
‾
,
X
i
−
X
‾
)
=
0
\quad 1.\quad E[\overline X] = \mu \\ \quad 2.\quad Var(\overline X) = \frac{\sigma^2}{n} \\ \quad 3.\quad Cov(\overline X,\ X_i - \overline X) = 0
1.E[X]=μ2.Var(X)=nσ23.Cov(X, Xi−X)=0
累积分布函数
F
X
+
Y
F_{X+Y}
FX+Y: X, Y连续,
F
X
+
Y
(
a
)
=
P
{
X
+
Y
≤
a
}
=
∫
∫
x
+
y
≤
a
f
(
x
)
g
(
y
)
d
y
=
∫
−
∞
∞
∫
−
∞
a
−
y
f
(
x
)
g
(
y
)
d
x
d
y
=
∫
−
∞
∞
(
∫
−
∞
a
−
y
f
(
x
)
d
x
)
g
(
y
)
d
y
=
∫
−
∞
∞
F
X
(
a
−
y
)
g
(
y
)
d
y
\begin{aligned} F_{X + Y}(a) & = P\{X + Y \leq a\} \\ & = \int \int \limits_{x + y \leq a} f(x) g(y) dy \\ & = \int \limits^{\infty}_{- \infty} \int \limits^{a - y}_{- \infty} f(x) g(y) dxdy \\ & = \int \limits^{\infty}_{- \infty} \left(\int \limits^{a - y}_{- \infty} f(x) dx \right) g(y) dy \\ & = \int \limits^\infty_{- \infty} F_X(a - y) g(y) dy \end{aligned}
FX+Y(a)=P{X+Y≤a}=∫x+y≤a∫f(x)g(y)dy=−∞∫∞−∞∫a−yf(x)g(y)dxdy=−∞∫∞⎝⎛−∞∫a−yf(x)dx⎠⎞g(y)dy=−∞∫∞FX(a−y)g(y)dy
概率密度
f
X
+
Y
(
a
)
f_{X + Y}(a)
fX+Y(a)
f
X
+
Y
(
a
)
=
d
d
a
F
(
X
+
Y
)
(
a
)
=
d
d
a
∫
−
∞
∞
F
X
(
a
−
y
)
g
(
y
)
d
y
=
∫
−
∞
∞
d
d
a
F
X
(
a
−
y
)
g
(
y
)
d
y
=
∫
−
∞
∞
f
(
a
−
y
)
g
(
y
)
d
y
\begin{aligned} f_{X + Y}(a) & = \frac{d}{da} F_{(X + Y)}(a) \\ & = \frac{d}{da} \int \limits^\infty_{- \infty} F_X(a - y) g(y) dy \\ & = \int \limits^\infty_{- \infty} \frac{d}{da} F_X(a - y) g(y) dy \\ & = \int \limits^\infty_{- \infty}f(a - y)g(y) dy \end{aligned}
fX+Y(a)=dadF(X+Y)(a)=dad−∞∫∞FX(a−y)g(y)dy=−∞∫∞dadFX(a−y)g(y)dy=−∞∫∞f(a−y)g(y)dy
例如:
随机变量的函数的联合概率分布:
1.
y
1
=
g
1
(
x
1
,
x
2
)
和
y
2
=
g
2
(
x
1
,
x
2
)
y_1 = g_1(x_1, x_2)和y_2 = g_2(x_1, x_2)
y1=g1(x1,x2)和y2=g2(x1,x2)可以唯一解出
x
1
x_1
x1和
x
2
x_2
x2,利用
y
1
y_1
y1和
y
2
y_2
y2给出
x
1
=
h
1
(
y
1
,
y
2
)
和
x
2
=
h
2
(
y
1
,
y
2
)
x_1 = h_1(y_1, y_2)和x_2 = h_2(y_1, y_2)
x1=h1(y1,y2)和x2=h2(y1,y2)
2.
g
1
和
g
2
g_1和g_2
g1和g2在所有的点
(
x
1
,
x
2
)
(x_1, x_2)
(x1,x2)上有连续的偏导数,且
J
(
x
1
,
x
2
)
=
∣
d
g
1
d
x
1
d
g
1
d
x
2
d
g
2
d
x
1
d
g
2
d
x
2
∣
≠
0
J(x_1, x_2) = \begin{vmatrix} \frac{d g_1}{d x_1} & \frac{d g_1}{d x_2} \\ \\ \frac{d g_2}{d x_1} & \frac{d g_2}{d x_2} \end{vmatrix} \neq 0
J(x1,x2)=∣∣∣∣∣∣dx1dg1dx1dg2dx2dg1dx2dg2∣∣∣∣∣∣=0
在这两个条件下,
Y
1
和
Y
2
Y_1和Y_2
Y1和Y2的联合密度函数为
f
Y
1
,
Y
2
(
y
1
,
y
2
)
=
f
X
1
,
X
2
(
x
1
,
x
2
)
∣
J
(
x
1
,
x
2
)
∣
−
1
f_{Y_1, Y_2}(y_1, y_2) = f_{X_1, X_2}(x_1, x_2) \ |J(x_1, x_2)|^{-1}
fY1,Y2(y1,y2)=fX1,X2(x1,x2) ∣J(x1,x2)∣−1
其中
x
1
=
h
1
(
y
1
,
y
2
)
,
x
2
=
h
2
(
y
1
,
y
2
)
x_1 = h_1(y_1, y_2),\ x_2 = h_2(y_1, y_2)
x1=h1(y1,y2), x2=h2(y1,y2)
例如:
从上面那个例题中,我们可以发现,
X
+
Y
和
X
/
(
X
+
Y
)
X + Y和X / (X + Y)
X+Y和X/(X+Y)是独立的,而且
X
+
Y
X + Y
X+Y有参数为
(
α
+
β
,
λ
)
(\alpha + \beta,\ \lambda)
(α+β, λ)的伽马分布,而
X
/
(
X
+
Y
)
X / (X + Y)
X/(X+Y)有密度函数
f
V
(
v
)
=
Γ
(
α
+
β
)
Γ
(
α
)
Γ
(
β
)
v
α
−
1
(
1
−
v
)
β
−
1
,
0
<
v
<
1
f_V(v) = \frac{\Gamma(\alpha + \beta)}{\Gamma(\alpha)\ \Gamma(\beta)} v^{\alpha - 1} (1 - v)^{\beta - 1},0 < v < 1
fV(v)=Γ(α) Γ(β)Γ(α+β)vα−1(1−v)β−1,0<v<1
联合概率分布个人理解是:
有一个由随机变量
X
,
Y
,
⋯
X, Y, \cdots
X,Y,⋯组成的概率空间,然后通过某种映射方式将
X
,
Y
,
⋯
X, Y, \cdots
X,Y,⋯映射到
U
,
V
,
⋯
U, V, \cdots
U,V,⋯,然后形成一个由
U
,
V
,
⋯
U, V, \cdots
U,V,⋯组成的新的概率空间。
贝塔密度(以
(
α
,
β
)
(\alpha,\ \beta)
(α, β)为参数):
Γ
(
α
+
β
)
Γ
(
α
)
Γ
(
β
)
v
α
−
1
(
1
−
v
)
β
−
1
,
0
<
v
<
1
\frac{\Gamma(\alpha + \beta)}{\Gamma(\alpha)\ \Gamma(\beta)} v^{\alpha - 1} (1 - v)^{\beta - 1},0 < v < 1
Γ(α) Γ(β)Γ(α+β)vα−1(1−v)β−1,0<v<1
5. 矩母函数
ϕ
(
t
)
=
E
[
e
t
X
]
=
{
∑
x
e
t
X
p
(
x
)
X
离
散
∫
−
∞
∞
e
t
x
f
(
x
)
d
x
X
连
续
\phi (t) = E[e^{tX}] = \begin{cases} \sum_x e^{tX} p(x) \quad X离散 \\ \\ \int^\infty_{-\infty} e^{tx} f(x) dx \quad X连续 \end{cases}
ϕ(t)=E[etX]=⎩⎪⎨⎪⎧∑xetXp(x)X离散∫−∞∞etxf(x)dxX连续
ϕ
(
t
)
\phi(t)
ϕ(t)称为矩母函数,因为X的所有矩都能有
ϕ
(
t
)
\phi(t)
ϕ(t)求微分(求导)得到
矩: 对变量分布和形态特点的一组度量
ϕ
(
n
)
(
0
)
=
E
[
X
n
]
\phi^{(n)}(0) = E[X^n]
ϕ(n)(0)=E[Xn],即矩母函数的n阶导在t=0时等于
E
[
X
n
]
E[X^n]
E[Xn]
期望:
ϕ
′
(
0
)
\phi^\prime(0)
ϕ′(0)
方差:
ϕ
′
′
(
0
)
−
[
ϕ
′
(
0
)
]
2
\phi^{\prime \prime}(0) - [\phi^\prime(0)]^2
ϕ′′(0)−[ϕ′(0)]2
ϕ
X
+
Y
(
t
)
=
E
[
e
t
(
X
+
Y
)
]
=
E
[
e
t
X
e
t
Y
]
=
E
[
e
t
X
]
E
[
e
t
Y
]
=
ϕ
X
(
t
)
ϕ
Y
(
t
)
\phi_{X+Y}(t) = E[e^{t(X+Y)}] = E[e^{tX}e^{tY}] = E[e^{tX}]E[e^{tY}] = \phi_X(t)\phi_Y(t)
ϕX+Y(t)=E[et(X+Y)]=E[etXetY]=E[etX]E[etY]=ϕX(t)ϕY(t)
独立随机变量和的矩母函数正是单个矩母函数的乘积
ϕ
(
t
1
,
⋯
,
t
n
)
=
E
[
e
(
t
1
X
1
+
⋯
+
t
n
X
n
)
]
\phi(t_1, \cdots, t_n) = E[e^{(t_1 X_1 + \cdots + t_n X_n)}]
ϕ(t1,⋯,tn)=E[e(t1X1+⋯+tnXn)]
矩母函数唯一地确定了分布,即随机变量的矩母函数和分布函数之间存在一一对应。
拉普拉斯变换:
g
(
t
)
=
ϕ
(
−
t
)
=
E
[
e
−
t
X
]
g(t) = \phi(-t) = E[e^{-tX}]
g(t)=ϕ(−t)=E[e−tX],即拉普拉斯变换在t处的赋值正是矩母函数在-t处的赋值。
拉普拉斯变换永远在0与1之间
有同样拉普拉斯变换的非负随机变量有同样的分布
样本方差: 假定
X
1
,
⋯
,
X
n
X_1, \cdots, X_n
X1,⋯,Xn是独立同分布随机变量,每个具有均值
μ
\mu
μ和方差
σ
2
\sigma^2
σ2,则样本方差为
S
2
=
∑
i
=
1
n
(
X
i
−
X
‾
)
2
n
−
1
S^2 = \sum^n_{i = 1} \frac{(X_i - \overline{X})^2}{n - 1}
S2=i=1∑nn−1(Xi−X)2
∑
i
=
1
n
(
X
i
−
X
‾
)
2
=
∑
i
=
1
n
(
X
i
−
μ
)
2
−
n
(
X
‾
−
μ
)
2
\sum^n_{i = 1} (X_i - \overline{X})^2 = \sum^n_{i = 1} (X_i - \mu)^2 - n(\overline{X} - \mu)^2
∑i=1n(Xi−X)2=∑i=1n(Xi−μ)2−n(X−μ)2
E
(
S
2
)
=
σ
2
E(S^2) = \sigma^2
E(S2)=σ2
卡方随机变量: 若
Z
1
,
⋯
,
Z
n
Z_1, \cdots, Z_n
Z1,⋯,Zn是独立的标准正态随机变量,那么随机变量
∑
i
=
1
n
Z
i
2
\sum^n_{i = 1} Z^2_i
∑i=1nZi2称为具有自由度n的卡方随机变量
如果
X
1
,
,
⋯
,
X
n
X_1,, \cdots, X_n
X1,,⋯,Xn是独立同分布的正态随机变量,具有均值
μ
\mu
μ和方差
σ
2
\sigma^2
σ2,那么样本均值
X
‾
\overline{X}
X与样本方差
S
2
S^2
S2是独立的。
X
‾
\overline{X}
X是正态随机变量,具有均值
μ
\mu
μ和方差
σ
2
n
\frac{\sigma^2}{n}
nσ2,
(
n
−
1
)
S
2
σ
2
\frac{(n - 1)S^2}{\sigma^2}
σ2(n−1)S2是具有n-1个自由度的卡方随机变量
6. 发生事件数的分布
7. 极限定理
马尔可夫不等式: X是只取非负值的随机变量,那么任意a > 0,有
P
{
X
≥
a
}
≤
E
[
X
]
a
P\{X \ge a\} \leq \frac{E[X]}{a}
P{X≥a}≤aE[X]
证明:
切比雪夫不等式: 如果X是具有均值
μ
\mu
μ和方差
σ
2
\sigma^2
σ2的随机变量,那么对于任意k>0,有
P
{
∣
X
−
μ
∣
≥
k
}
≤
σ
2
k
2
P\{|X - \mu| \geq k\} \leq \frac{\sigma^2}{k^2}
P{∣X−μ∣≥k}≤k2σ2
在只有概率分布的均值或者均值和方差已知时,它们使我们可以推得所求概率的上界
例如:
强大数定律: 设
X
1
,
X
2
,
⋯
X_1, X_2, \cdots
X1,X2,⋯是一列独立同分布的随机变量,令
E
[
X
i
]
=
μ
E[X_i] = \mu
E[Xi]=μ,那么当
n
→
∞
n \rightarrow \infty
n→∞时以概率1有
X
1
+
X
2
+
⋯
+
X
n
n
→
μ
\frac{X_1 + X_2 + \cdots + X_n}{n} \rightarrow \mu
nX1+X2+⋯+Xn→μ
即一列独立同分布的随机变量的平均值以概率1收敛到这个分布的均值
中心极限定理: 假定
X
1
,
X
2
,
⋯
X_1, X_2, \cdots
X1,X2,⋯是一列独立同分布的随机变量,每个具有均值
μ
\mu
μ和方差
σ
2
\sigma^2
σ2,那么当
n
→
∞
n \rightarrow \infty
n→∞时
X
1
+
X
2
+
⋯
+
X
n
−
n
μ
σ
n
\frac{X_1 + X_2 + \cdots + X_n - n \mu}{\sigma \sqrt{n}}
σnX1+X2+⋯+Xn−nμ的分布趋于标准正态分布,即
n
→
∞
,
P
{
X
1
+
⋯
+
X
n
−
n
μ
σ
n
≤
a
}
→
1
2
π
∫
−
∞
a
e
−
x
2
2
d
x
n \rightarrow \infty,P\{\frac{X_1 + \cdots + X_n - n\mu}{\sigma \sqrt{n}} \leq a\} \rightarrow \frac{1}{\sqrt{2\pi}}\int^a_{-\infty} e^{-\frac{x^2}{2}}dx
n→∞,P{σnX1+⋯+Xn−nμ≤a}→2π1∫−∞ae−2x2dx
例如:
8. 随机过程
一个随机过程 { X ( t ) , t ∈ T } \{X(t), t \in T\} {X(t),t∈T}是随机变量的一个集合。集合T称为此过程的指标集。当T是可数集时,随机过程称为离散时间过程;当T是一个实数空间时,随机过程称为连续时间过程。
随机过程的本质有两个要点:一是随机,二是过程;随机说明任何时候结果都存在不确定性,即分布函数(或者概率密度函数);过程体现的是时间;在时间t时,随机变量服从某一分布,另一时刻随机变量服从某一分布。如下图所示
参考资料:
- https://blog.csdn.net/ningyanggege/article/details/88549654
- 应用随机过程概率模型导论 第11版