随机变量
X
X
X服从参数
(
α
,
β
)
(\alpha,\beta)
(α,β)的Gamma分布，则其概率密度函数(pdf)可以表示为
f
(
x
)
=
β
α
Γ
(
α
)
x
α
−
1
e
−
β
x
,
(
x
>
0
)
(1)
f(x)=\frac{\beta^\alpha}{\Gamma(\alpha)}x^{\alpha-1}e^{-\beta x},(x > 0)\tag{1}
f(x)=Γ(α)βαxα−1e−βx,(x>0)(1)
其对应的矩母函数为
M
x
(
t
)
=
(
1
+
t
/
β
)
−
α
{\rm M}_x(t)=(1+t/\beta)^{-\alpha}
Mx(t)=(1+t/β)−α，所以若
X
1
∽
G
(
α
1
,
β
1
)
,
X
2
∽
G
(
α
2
,
β
2
)
,
X
=
X
1
+
X
2
X_1 \backsim G(\alpha_1,\beta_1),X_2 \backsim G(\alpha_2,\beta_2),X=X_1+X_2
X1∽G(α1,β1),X2∽G(α2,β2),X=X1+X2，且
β
1
=
β
2
=
β
\beta_1=\beta_2=\beta
β1=β2=β，则由矩母函数可以很容易得到
M
x
(
t
)
=
M
x
1
(
t
)
M
x
2
(
t
)
=
(
1
+
t
/
β
)
−
(
α
1
+
α
2
)
{\rm M}_x(t)={\rm M}_{x_1}(t){\rm M}_{x_2}(t)=(1+t/\beta)^{-(\alpha_1+\alpha_2)}
Mx(t)=Mx1(t)Mx2(t)=(1+t/β)−(α1+α2)，所以
X
∽
G
(
α
1
+
α
2
,
β
)
X \backsim G(\alpha_1+\alpha_2,\beta)
X∽G(α1+α2,β)，该性质称为Gamma分布的可加性，其更具体的推导过程可以查看《不同方法推导Gamma分布可加性产生的矛盾》。对于更一般的情况，即两个随机变量服从参数不同的Gamma分布，则这两个随机变量和的随机变量的矩母函数可以表示
M
x
(
t
)
=
M
x
1
(
t
)
M
x
2
(
t
)
=
(
1
+
t
/
β
1
)
−
α
1
(
1
+
t
/
β
2
)
−
α
2
{\rm M}_x(t)={\rm M}_{x_1}(t){\rm M}_{x_2}(t)=(1+t/\beta_1)^{-\alpha_1}(1+t/\beta_2)^{-\alpha_2}
Mx(t)=Mx1(t)Mx2(t)=(1+t/β1)−α1(1+t/β2)−α2，此时不再服从上述可加性性质。本文将就上述
α
1
,
α
2
∈
Z
+
\alpha_1,\alpha_2 \in \Bbb Z^+
α1,α2∈Z+情况下的两个Gamma分布随机变量的和的pdf和cdf进行推导。
根据概率论相关知识，我们知道，两个随机变量的和的pdf等于各自pdf的卷积，因此
f
x
(
x
)
=
∫
−
∞
∞
f
x
1
(
y
)
f
x
2
(
x
−
y
)
d
y
(2)
\begin{equation} \begin{aligned} f_x(x)=\int_{-\infty}^\infty f_{x_1}(y)f_{x_2}(x-y)dy \end{aligned} \end{equation}\tag{2}
fx(x)=∫−∞∞fx1(y)fx2(x−y)dy(2)
易知，仅当
y
>
0
,
x
−
y
>
0
y>0,x-y>0
y>0,x−y>0，即
y
>
0
,
y
<
x
y>0,y<x
y>0,y<x时，上述积分的被积函数不等于0，将相关表达式带入(2)可以得到
f
x
(
x
)
=
β
1
α
1
β
2
α
2
Γ
(
α
1
)
Γ
(
α
2
)
∫
0
x
y
α
1
−
1
e
−
β
1
y
(
x
−
y
)
α
2
−
1
e
−
β
2
(
x
−
y
)
d
y
=
β
1
α
1
β
2
α
2
Γ
(
α
1
)
Γ
(
α
2
)
e
−
β
2
x
∫
0
x
y
α
1
−
1
(
x
−
y
)
α
2
−
1
e
−
(
β
1
−
β
2
)
y
d
y
(3)
\begin{equation} \begin{aligned} f_x(x)&=\frac{\beta_1^{\alpha_1}\beta_2^{\alpha_2}}{\Gamma(\alpha_1)\Gamma(\alpha_2)} \int_0^x y^{\alpha_1-1}e^{-\beta_1 y}(x-y)^{\alpha_2-1}e^{-\beta_2(x-y)}dy\\ &=\frac{\beta_1^{\alpha_1}\beta_2^{\alpha_2}}{\Gamma(\alpha_1)\Gamma(\alpha_2)} e^{-\beta_2 x} \int_0^x y^{\alpha_1-1}(x-y)^{\alpha_2-1}e^{-(\beta_1-\beta_2)y}dy \end{aligned} \end{equation}\tag{3}
fx(x)=Γ(α1)Γ(α2)β1α1β2α2∫0xyα1−1e−β1y(x−y)α2−1e−β2(x−y)dy=Γ(α1)Γ(α2)β1α1β2α2e−β2x∫0xyα1−1(x−y)α2−1e−(β1−β2)ydy(3)
由二项展开定理有:
(
x
−
y
)
α
2
−
1
=
∑
i
=
0
α
2
−
1
(
−
1
)
α
2
−
1
−
i
(
α
2
−
1
i
)
x
i
y
α
2
−
1
−
i
(x-y)^{\alpha_2-1}=\sum_{i=0}^{\alpha_2-1}(-1)^{\alpha_2-1-i} \begin{pmatrix}\alpha_2-1\\ i\end{pmatrix}x^i y^{\alpha_2-1-i}
(x−y)α2−1=∑i=0α2−1(−1)α2−1−i(α2−1i)xiyα2−1−i，将其代入(3)并化简有
f
x
(
x
)
=
β
1
α
1
β
2
α
2
Γ
(
α
1
)
Γ
(
α
2
)
e
−
β
2
x
∑
i
=
0
α
2
−
1
(
−
1
)
α
2
−
1
−
i
(
α
2
−
1
i
)
x
i
∫
0
x
y
α
1
+
α
2
−
i
−
2
e
−
(
β
1
−
β
2
)
d
y
(4)
\begin{equation} \begin{aligned} f_x(x)&=\frac{\beta_1^{\alpha_1}\beta_2^{\alpha_2}}{\Gamma(\alpha_1)\Gamma(\alpha_2)} e^{-\beta_2 x} \sum_{i=0}^{\alpha_2-1}(-1)^{\alpha_2-1-i} \begin{pmatrix}\alpha_2-1\\ i\end{pmatrix}x^i \int_0^x y^{\alpha_1+\alpha_2-i-2}e^{-(\beta_1-\beta_2)}dy \end{aligned} \end{equation}\tag{4}
fx(x)=Γ(α1)Γ(α2)β1α1β2α2e−β2xi=0∑α2−1(−1)α2−1−i(α2−1i)xi∫0xyα1+α2−i−2e−(β1−β2)dy(4)
上述积分式的求导，需要用到下面的定理[1]：
∫
0
μ
x
m
e
−
t
x
d
x
=
m
!
t
−
(
m
+
1
)
[
1
−
∑
k
=
0
m
t
k
k
!
μ
k
e
−
t
μ
]
μ
→
∞
→
∫
0
μ
x
m
e
−
t
x
d
x
=
m
!
t
−
(
m
+
1
)
(5)
\begin{equation} \begin{aligned} \int_0^\mu{x^m e^{-tx}dx}=m!t^{-(m+1)}[1-\sum_{k=0}^m{\frac{t^k}{k!} \mu^k e^{-t\mu}}] \quad\underrightarrow{\mu \rightarrow \infty} \quad \int_0^\mu{x^m e^{-tx}dx}=m!t^{-(m+1)} \end{aligned} \end{equation}\tag{5}
∫0μxme−txdx=m!t−(m+1)[1−k=0∑mk!tkμke−tμ]μ→∞∫0μxme−txdx=m!t−(m+1)(5)
所以
∫
0
x
y
α
1
+
α
2
−
i
−
2
e
−
(
β
1
−
β
2
)
d
y
=
(
α
1
+
α
2
−
i
−
2
)
!
(
β
1
−
β
2
)
−
(
α
1
+
α
2
−
i
−
1
)
[
1
−
∑
k
=
0
α
1
+
α
2
−
i
−
2
(
β
1
−
β
2
)
k
k
!
x
k
e
−
(
β
1
−
β
2
)
x
]
(6)
\int_0^x y^{\alpha_1+\alpha_2-i-2}e^{-(\beta_1-\beta_2)}dy=(\alpha_1+\alpha_2-i-2)!(\beta_1-\beta_2)^{-(\alpha_1+\alpha_2-i-1)}[1-\sum_{k=0}^{\alpha_1+\alpha_2-i-2}{\frac{(\beta_1-\beta_2)^k}{k!} x^k e^{-(\beta_1-\beta_2) x}}]\tag{6}
∫0xyα1+α2−i−2e−(β1−β2)dy=(α1+α2−i−2)!(β1−β2)−(α1+α2−i−1)[1−k=0∑α1+α2−i−2k!(β1−β2)kxke−(β1−β2)x](6)
将(6)代入(4)整理得到
f
x
(
x
)
=
A
0
∑
i
=
0
α
2
−
1
A
1
(
i
)
[
x
i
e
−
β
2
x
−
∑
k
=
0
α
1
+
α
2
−
i
−
2
(
β
1
−
β
2
)
k
k
!
x
k
+
i
e
−
β
1
x
]
(7)
\begin{equation} \begin{aligned} f_x(x)=A_0\sum_{i=0}^{\alpha_2-1}A_1(i)[x^i e^{-\beta_2 x}-\sum_{k=0}^{\alpha_1+\alpha_2-i-2}{\frac{(\beta_1-\beta_2)^k}{k!} x^{k+i} e^{-\beta_1 x}}] \end{aligned} \end{equation}\tag{7}
fx(x)=A0i=0∑α2−1A1(i)[xie−β2x−k=0∑α1+α2−i−2k!(β1−β2)kxk+ie−β1x](7)
其中
A
0
=
Δ
A
0
(
α
1
,
α
2
,
β
1
,
β
2
)
=
β
1
α
1
β
2
α
2
Γ
(
α
1
)
Γ
(
α
2
)
A
1
(
i
)
=
Δ
A
1
(
α
1
,
α
2
,
β
1
,
β
2
,
i
)
=
(
−
1
)
α
2
−
1
−
i
(
α
2
−
1
i
)
(
α
1
+
α
2
−
i
−
2
)
!
(
β
1
−
β
2
)
−
(
α
1
+
α
2
−
i
−
1
)
(8)
\begin{equation} \begin{aligned} A_0 &\overset{\Delta}{=} A_0(\alpha_1,\alpha_2,\beta_1,\beta_2)=\frac{\beta_1^{\alpha_1}\beta_2^{\alpha_2}}{\Gamma(\alpha_1)\Gamma(\alpha_2)}\\ A_1(i)&\overset{\Delta}{=}A_1(\alpha_1,\alpha_2,\beta_1,\beta_2,i)=(-1)^{\alpha_2-1-i} \begin{pmatrix}\alpha_2-1\\ i\end{pmatrix}(\alpha_1+\alpha_2-i-2)!(\beta_1-\beta_2)^{-(\alpha_1+\alpha_2-i-1)} \end{aligned} \end{equation}\tag{8}
A0A1(i)=ΔA0(α1,α2,β1,β2)=Γ(α1)Γ(α2)β1α1β2α2=ΔA1(α1,α2,β1,β2,i)=(−1)α2−1−i(α2−1i)(α1+α2−i−2)!(β1−β2)−(α1+α2−i−1)(8)
X
X
X的cdf可以表示为其pdf的积分，因此
F
x
(
x
)
=
∫
0
x
f
x
(
x
)
d
x
=
I
0
−
I
1
(9)
\begin{equation} \begin{aligned} F_x(x)=\int_0^x f_x(x)dx=I_0-I_1 \end{aligned} \end{equation}\tag{9}
Fx(x)=∫0xfx(x)dx=I0−I1(9)
其中
$$
\begin{equation}
\begin{aligned}
I_0&=A_0\sum_{i=0}{\alpha_2-1}A_1(i)\int_0x x^i e^{-\beta_2 x}dx\
&=A_0\sum_{i=0}^{\alpha_2-1}A_1(i) i!\beta_2{-(i+1)}[1-\sum_{k=0}i{\frac{\beta_2^k}{k!} x^k e^{-\beta_2 x}}]\
I_1&=A_0\sum_{i=0}^{\alpha_2-1}A_1(i)\sum_{k=0}^{\alpha_1+\alpha_2-i-2}{\frac{(\beta_1-\beta_2)^k}{k!} \int_0^x x^{k+i} e^{-\beta_1 x}}dx\\
&=A_0\sum_{i=0}^{\alpha_2-1}A_1(i)\sum_{k=0}^{\alpha_1+\alpha_2-i-2}\frac{(\beta_1-\beta_2)^k}{k!}(k+i)!\beta_1^{-(k+i+1)}[1-\sum_{j=0}^{k+i}{\frac{\beta_1^j}{j!} x^j e^{-\beta_1 x}}]
\end{aligned}
\end{equation}\tag{10}
将
(
10
)
带入
(
9
)
并整理得到
将(10)带入(9)并整理得到
将(10)带入(9)并整理得到
\begin{equation}
\begin{aligned}
F_x(x)&=A_0\sum_{i=0}{\alpha_2-1}A_1(i)[i!\beta_2{-(i+1)}-\sum_{k=0}{\alpha_1+\alpha_2-i-2}\frac{(\beta_1-\beta_2)k}{k!}(k+i)!\beta_1^{-(k+i+1)}]\
&-A_0\sum_{i=0}{\alpha_2-1}A_1(i)[i!\sum_{k=0}i{\frac{\beta_2^{-(i-k+1)}}{k!} x^k e^{-\beta_2 x}}-\sum_{k=0}{\alpha_1+\alpha_2-i-2}\frac{(\beta_1-\beta_2)k}{k!}(k+i)!\sum_{j=0}{k+i}{\frac{\beta_1{-(k+i-j+1)}}{j!} x^j e^{-\beta_1 x}}]
\end{aligned}
\end{equation}\tag{11}
通过数值仿真发现
(
11
)
式中的前一部分满足
通过数值仿真发现(11)式中的前一部分满足
通过数值仿真发现(11)式中的前一部分满足
A_0\sum_{i=0}{\alpha_2-1}A_1(i)[i!\beta_2{-(i+1)}-\sum_{k=0}{\alpha_1+\alpha_2-i-2}\frac{(\beta_1-\beta_2)k}{k!}(k+i)!\beta_1^{-(k+i+1)}]\equiv 1
$$
但是目前还不知道如何证明？显然此时pdf和cdf的表达式均很复杂，也有可能还没有化简到最简形式，但是暂时不知道怎样化简下去了。
[1] Table of Integrals, Series, and Products
1696
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
