概率论第六周作业
习题 3 T14
若 \(f_1(x),f_2(x),f_3(x)\) 是对应于分布函数 \(F_1(x),F_2(x),F_3(x)\) 的密度函数,证明对于一切 \(\alpha (-1 < \alpha < 1)\) ,下列函数是密度函数,且具有相同的边际密度函数 \(f_1(x),f_2(x),f_3(x)\) :
$$ \begin{aligned} &f_\alpha(x_1,x_2,x_3) \ & = f_1(x_1)f_2(x_2)f_3(x_3)\left\lbrace 1+ \alpha [2F_1(x_1)-1][2F_2(x_2)-1][2F_3(x_3)-1] \right\rbrace \end{aligned} $$
首先验证非负性:易知
\[
0 \leqslant f_1(x_1)f_2(x_2)f_3(x_3) \leqslant 1
\]
由 \(-1 \leqslant 2F_i(x_i)-1 \leqslant 1\) ,可知
\[
0 \leqslant 1+\alpha [2F_1(x_1)-1][2F_2(x_2)-1][2F_3(x_3)-1] \leqslant 1
\]
因此 \(f_\alpha(x_1,x_2,x_3) \geqslant 0\) 成立.
然后验证积分为 \(1\) ,根据表达式易知,\(f_\alpha( - \infty,-\infty,-\infty) = 0, f_\alpha(+\infty,+\infty,+\infty)=1\) . 且可计算:
\[
\begin{aligned}
&\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty} f_\alpha(x_1,x_2,x_3)\mathrm{d}x_1 \mathrm{d}x_2 \mathrm{d}x_3 \\
& = \int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty} f_1(x_1)f_2(x_2)f_3(x_3)\mathrm{d}x_1 \mathrm{d}x_2 \mathrm{d}x_3 \\
& + \alpha\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty} f_1(x_1)f_2(x_2)f_3(x_3) [2F_1(x_1)-1][2F_2(x_2)-1][2F_3(x_3)-1]\mathrm{d}x_1 \mathrm{d}x_2\mathrm{d}x_3 \\
& = 1+ 0 = 1
\end{aligned}
\]
其中第二个等号如下推导:
\[
\begin{aligned}
& \int_{- \infty}^{+\infty} [2F_i(x_i)-1] f_i(x_i) \mathrm{d}x_2 \\
& = \int_{- \infty}^{+\infty} [2F_i(x_i)-1]\mathrm{d} F_i(x_i) \\
& = \int_0^1 (2t-1)\mathrm{d} t \\
& = 0
\end{aligned}
\]
用到了分布密度函数的定义以及 \(F_i(+ \infty) = 1, F_i(- \infty) = 0\) . \(i=1,2,3\) . 因此该函数为密度函数.
再计算
\[
\begin{aligned}
&\int_{-\infty}^{x_1}\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty} f_\alpha(u,x_2,x_3)\mathrm{d}u \mathrm{d}x_2 \mathrm{d}x_3 \\
& = \int_{-\infty}^{x_1}\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty} f_1(u)f_2(x_2)f_3(x_3)\mathrm{d}u \mathrm{d}x_2 \mathrm{d}x_3 \\
& + \alpha\int_{-\infty}^{x_1}\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty} f_1(u)f_2(x_2)f_3(x_3) [2F_1(u)-1][2F_2(x_2)-1][2F_3(x_3)-1]\mathrm{d}u \mathrm{d}x_2\mathrm{d}x_3 \\
& = \int_{- \infty}^{x_1}f_1(u)\mathrm{d}u+ 0 = F_1(x_1)
\end{aligned}
\]
因此 \(f_\alpha(x_1,x_2,x_3)\) 有边际分布密度函数 \(f_1(x_1)\) ,同理 \(f_2(x_2),f_3(x_3)\) 也为边际分布密度函数. \(\square\)
习题 3 T15
若 \((\xi,\eta)\) 的联合概率分布为
$$ \begin{array}{c|ccc} \eta \backslash \xi & -1 & 0 & 1 \\ \hline -1 & a & 0 & 0.2 \\ 0 & 0.1 & b & 0.1 \\ 1 & 0 & 0.2 & c \end{array} $$
且 \(P \left\lbrace \xi \eta \neq 0 \right\rbrace = 0.4 , P \left\lbrace \eta \leqslant 0 \mid \xi \leqslant 0 \right\rbrace = \dfrac{2}{3}\) ,试求:
- \(a,b,c\) 之值;
- \(\xi\) 和 \(\eta\) 的边际概率分布;
- \(\xi+\eta\) 的概率分布.
(1) 首先整体求和为 \(1\) :
\[
a+b+c+0.6 = 1 \tag{2.1}
\]
然后考虑 \(P \left\lbrace \xi \eta \neq 0 \right\rbrace = 0.4\) ,也就是 \(P \left\lbrace \eta \neq 0 \land \xi \neq 0 \right\rbrace = 0.4\) :
\[
a+c+0.2 = 0.4 \tag{2.2}
\]
根据
\[
P \left\lbrace \eta\leqslant 0\mid \xi \leqslant 0 \right\rbrace = \dfrac{P \left\lbrace \eta \leqslant 0 \land \xi \leqslant 0 \right\rbrace}{P \left\lbrace \xi \leqslant 0 \right\rbrace} = \dfrac{a+b+0.1}{a+b+0.3}
\]
故
\[
\dfrac{a+b+0.1}{a+b+0.3} = \frac{2}{3}\tag{2.3}
\]
根据 \((2.1),(2.2),(2.3)\) 可得到
\[
\begin{cases}
a = 0.1 \\
b= 0.2 \\
c = 0.1
\end{cases}
\]
(2) 根据 (1) 有
\[
\begin{array}{c|ccc}
\eta \backslash \xi & -1 & 0 & 1 \\ \hline
-1 & 0.1 & 0 & 0.2 \\
0 & 0.1 & 0.2 & 0.1 \\
1 & 0 & 0.2 & 0.1
\end{array}
\]
设 \((\xi,\eta)\) 的分布密度函数为 \(p(x,y)\) ,此时固定一个变元,对令一个变元的概率求和即可得:
\(\xi\) 的边际概率分布:
\[
\begin{array}{|c|ccc|}
\hline
\xi & -1 & 0 & 1 \\
\hline
P & 0.2 & 0.4 & 0.4 \\
\hline
\end{array}
\]
\(\eta\) 的边际概率分布:
\[
\begin{array}{|c|ccc|}
\hline
\eta & -1 & 0 & 1 \\
\hline
P & 0.3 & 0.4 & 0.3 \\
\hline
\end{array}
\]
(3) 利用联合概率分布有:
\[
\begin{array}{|c|ccccc|}
\hline
\xi+ \eta & -2 & -1 & 0 & 1 & 2 \\
\hline
p_{\xi+ \eta} & 0.1 & 0.1 & 0.4 & 0.3 & 0.1 \\
\hline
\end{array}
\]
\(\square\)
习题 3 T18
设二维随机变量 \((\xi,\eta)\) 的联合密度为
$$ p(x,y) = \dfrac{1}{\Gamma(k_1)\Gamma(k_2)} x^{k_1-1}(y-x)^{k_2-1} \mathrm{e}^{-y} $$
\(k_1> 0,k_2 > 0, 0 < x \leqslant y < \infty\) ,试求 \(\xi\) 和 \(\eta\) 边际分布密度.
对于 \(\xi\) 的边际分布密度,根据 \(x,y\) 的范围,计算如下积分:
\[
\begin{aligned}
p_\xi (x) & = \int_{0}^{+\infty} \dfrac{1}{\Gamma(k_1)\Gamma(k_2)} x^{k_1-1}(y-x)^{k_2-1} \mathrm{e}^{-y} \mathrm{d} y \\
\end{aligned}
\]
仅需考虑如下积分:
\[
\begin{aligned}
\int_{0}^{+\infty} (y-x)^{k_2-1} \mathrm{e}^{-y}\mathrm{d}y = \mathrm{e}^{-x}\int_{0}^{+\infty} (y-x)^{k_2-1} \mathrm{e}^{-(y-x)}\mathrm{d}(y-x) = \mathrm{e}^{-x} \Gamma (k_2)
\end{aligned}
\]
有
\[
p_\xi(x) = \dfrac{1}{\Gamma(k_1)}x^{k_1-1}\mathrm{e}^{-x}
\]
对 \(\eta\) 有
\[
p_\eta (y) = \int_{0}^{y} \dfrac{1}{\Gamma(k_1)\Gamma(k_2)} x^{k_1-1}(y-x)^{k_2-1} \mathrm{e}^{-y} \mathrm{d} x
\]
那么由
\[
\begin{aligned}
& \int_0^{y} x^{k_1-1} (y-x)^{k_2-1} \mathrm{d}x= y^{k_1+k_2-1}\int_0^1 \left(\dfrac{y}{x}\right)^{k_1-1}\left(1- \frac{y}{x}\right)^{k_2-1}\mathrm{d} \dfrac{y}{x} \\ & = y^{k_1+k_2-1}\mathrm{B} ( k_1,k_2) = y^{k_1+k_2-1}\dfrac{\Gamma(k_1)\Gamma(k_2)}{\Gamma(k_1+k_2)}
\end{aligned}
\]
从而
\[
p_\eta (y) = y^{k_1+k_2-1}\mathrm{e}^{-y} \dfrac{1}{\Gamma(k_1+k_2)}
\]
\(\square\)
习题 3 T20
(1) 若 \((\xi,\eta)\) 的联合密度函数为
$$ f(x,y) = \begin{cases}4xy, & 0\leqslant x \leqslant 1, 0 \leqslant y \leqslant 1 \\ 0, & \text{Otherwise}.\end{cases} $$
问 \(\xi\) 和 \(\eta\) 是否相互独立?
(2) 若 \((\xi,\eta)\) 的联合密度函数为
$$ g(x,y) = \begin{cases}8xy, & 0\leqslant x \leqslant y, 0 \leqslant y \leqslant 1 \\ 0, & \text{Otherwise}.\end{cases} $$
问 \(\xi\) 和 \(\eta\) 是否相互独立?
(1) 首先计算边际分布密度函数:
\[
f_\xi(x) = \int_{0}^1 f(x,y)\mathrm{d}y = \int_0^1 4xy \mathrm{d}y = 2x
\]
\[
f_\eta(y) = \int_0^1 f(x,y)\mathrm{d}x = 2y
\]
从而 \(f(x,y) = 4xy = f_\xi(x)f_\eta(y)\) ,因此 \(\xi\) 和 \(\eta\) 相互独立.
(2) 计算边际分布密度函数:
\[
g_\xi(x) = \int_0^1 g(x,y)\mathrm{d}y = 4x
\]
\[
g_\eta(y) = \int_0^y g(x,y)\mathrm{d}x = 4y^3
\]
故 \(g_\xi(x)g_\eta(y) = 16xy^3 \neq g(x,y)\) .
习题 3 T21
若 \(\xi,\eta\) 相互独立,且以概率 \(\dfrac{1}{2}\) 取值 \(+1\) 和 \(-1\) ,令 \(\zeta = \xi \eta\) ,试证:\(\xi,\eta,\zeta\) 两两独立但是不相互独立.
根据 \(\xi,\eta\) 相互独立,有
\[
P \left\lbrace \zeta = 1 \right\rbrace = P \left\lbrace \xi=1 \right\rbrace P \left\lbrace \eta=1 \right\rbrace + P \left\lbrace \xi=-1 \right\rbrace P \left\lbrace \eta=-1 \right\rbrace = 0.5
\]
同理 \(P \left\lbrace \zeta=-1 \right\rbrace=0.5\) .
由于 \(\xi,\eta,\zeta\) 均为离散型随机变量,考虑各取值:
\[
\begin{cases}
P \left\lbrace \xi = 1, \zeta = 1 \right\rbrace = \dfrac{1}{4} = P \left\lbrace \xi=1 \right\rbrace P \left\lbrace \zeta=1 \right\rbrace \\
P \left\lbrace \xi = 1, \zeta = -1 \right\rbrace = \dfrac{1}{4} = P \left\lbrace \xi=1 \right\rbrace P \left\lbrace \zeta=-1 \right\rbrace \\
P \left\lbrace \xi = -1, \zeta = 1 \right\rbrace = \dfrac{1}{4} = P \left\lbrace \xi=-1 \right\rbrace P \left\lbrace \zeta=1 \right\rbrace \\
P \left\lbrace \xi = -1, \zeta = -1 \right\rbrace = \dfrac{1}{4} = P \left\lbrace \xi=-1 \right\rbrace P \left\lbrace \zeta=-1 \right\rbrace
\end{cases}
\]
因此 \(\xi,\zeta\) 相互独立,同理 \(\eta,\zeta\) 相互独立,从而两两独立成立.
但是由于
\[
P \left\lbrace \xi=1,\eta=1,\zeta=1 \right\rbrace = \frac{1}{4} \neq P \left\lbrace \xi=1 \right\rbrace P \left\lbrace \eta=1 \right\rbrace P \left\lbrace \zeta=1 \right\rbrace
\]
可知相互独立不成立. \(\square\)
习题 3 T22
设 \((\xi,\eta)\) 具有联合密度函数
$$ p(x,y) = \begin{cases}\dfrac{1+xy}{4}, & |x|<1,|y|<1 \\ 0, & \text{Otherwise}.\end{cases} $$
试证:\(\xi\) 和 \(\eta\) 不独立,但是 \(\xi^2\) 和 \(\eta^2\) 是相互独立的.
首先计算边际分布密度函数:
\[
p_\xi(x) = \int_{-1}^1 \dfrac{1+xy}{4}\mathrm{d}y = \frac{1}{2}
\]
\[
p_\eta(y) = \int_{-1}^1 \dfrac{1+xy}{4}\mathrm{d}x = \frac{1}{2}
\]
因此 \(p(x,y)\neq \dfrac{1}{4} = p_\xi(x)p_\eta(y)\) ,即 \(\xi\) 和 \(\eta\) 不相互独立.
再计算分布函数:
\[
\begin{aligned}
F(x,y) & = \int_{-1}^{x}\int_{-1}^y \dfrac{1+uv}{4}\mathrm{d}u \mathrm{d}v \\
& = \int_{-1}^x \left(\frac{1}{4}y + \frac{1}{4}+ \frac{1}{8}uy^2- \frac{1}{8}u\right)\mathrm{d}u \\
& = \frac{1}{4}(x+1)(y+1) + \frac{1}{16}(y^2-1)(x^2-1)
\end{aligned} \tag{22.1}
\]
计算
\[
P \left\lbrace \xi < \sqrt{x} \right\rbrace = \int_{-1}^{\sqrt{x}}\int_{-1}^1 \dfrac{1+uy}{4}\mathrm{d}y \mathrm{d}u= \frac{1}{2}\sqrt{x}+ \frac{1}{2}
\]
\[
P \left\lbrace -\sqrt{x} < \xi < \sqrt{x} \right\rbrace
= P \left\lbrace \xi < \sqrt{x} \right\rbrace - P \left\lbrace \xi \leqslant - \sqrt{x} \right\rbrace = \sqrt{x}
\]
从而 \(P \left\lbrace \xi^2 < x \right\rbrace = \sqrt{x}\) . 同理 \(P \left\lbrace \eta^2 < y \right\rbrace = \sqrt{y}\) .
而
\[
\begin{aligned}
P \left\lbrace \xi^2 < x, \eta^2 < y \right\rbrace & = P \left\lbrace - \sqrt{x} <\xi < \sqrt{x} , -\sqrt{ y} <\eta< \sqrt{y} \right\rbrace \\
& = P \left\lbrace \xi < \sqrt{x}, \eta < \sqrt{y} \right\rbrace - P \left\lbrace \xi \leqslant -\sqrt{x}, \eta< \sqrt{y} \right\rbrace \\ & -P \left\lbrace \xi< \sqrt{x}, \eta \leqslant - \sqrt{y} \right\rbrace + P \left\lbrace \xi \leqslant -\sqrt{x}, \eta \leqslant -\sqrt{y} \right\rbrace
\end{aligned}
\]
代入 \((22.1)\) 有
\[
P \left\lbrace \xi^2 < x,\eta^2< y \right\rbrace =\sqrt{x}\sqrt{y} = P \left\lbrace \xi^2 < x \right\rbrace P \left\lbrace \eta^2 < y \right\rbrace
\]
故 \(\eta^2\) 和 \(\xi^2\) 相互独立. \(\square\)