概率论第九周作业
习题四T3
设随机变量 \(\xi\) 只取非负整数值,其概率为 \(P \left\lbrace \xi=k \right\rbrace = \dfrac{a^k}{(1+a)^{k+1}},a>0\) 是常数,试求 \(E(\xi)\) 以及 \(D(\xi)\).
容易化为几何分布的形式:
\[
P \left\lbrace \xi = k \right\rbrace = \left(\dfrac{a}{1+a} \right)^k \dfrac{1}{1+a}, k=0,1,\cdots
\]
令 \(q = \dfrac{a}{1+a}\) ,对 \(E(\xi)\) ,有
\[
\begin{aligned}
E(\xi) & = (1-q)\sum\limits_{k=1}^\infty k q^k \\
& = q(1-q)\sum\limits_{k=1}^\infty kq^{k-1} \\
& = q(1-q)\dfrac{1}{(1-q)^2} \\
& = \dfrac{q}{1-q} = a
\end{aligned}
\]
利用 \(D(\xi) = E(\xi^2)-E(\xi)^2\) ,计算
\[
\begin{aligned}
E(\xi^2) & = (1-q)\sum\limits_{k=1}^\infty k^2 q^k \\
& = q(1-q)\sum\limits_{k=1}^\infty k^2 q^{k-1} \\
& = 2a^2+a
\end{aligned}
\]
从而 \(D(\xi) = 2a^2+a-a^2 = a^2+a\) . \(\square\)
习题四 T8
若随机变量 \(\xi\) 的分布函数为 \(F(x)\),试证:
$$ E(\xi) = \int_0^\infty [1-F(x)]\mathrm{d}x-\int_{-\infty}^0 F(x) \mathrm{d}x $$
特别地,若 \(\xi\) 取非负值,则
$$ E(\xi) = \int_0^{\infty} [1-F(x)] \mathrm{d}x $$
根据期望的存在性:
\[
\int_{-\infty}^{+\infty}|x| \mathrm{d}F(x) < +\infty
\]
那么,对 \(x F(x)\) ,有
\[
\int_{-\infty}^{-u} |x| \mathrm{d}F(x) \geqslant u [F(-u)-F(- \infty)] = u F(-u) \geqslant 0
\]
对于任意的 \(u\) 均成立.
习题四 T9
若随机变量 \(\xi\) 服从 Laplace 分布,其密度函数为
$$ p(x) = \dfrac{1}{2\lambda} \mathrm{e}^{-\frac{|x-\mu|}{\lambda}},x\in \mathbb{R},\lambda>0 $$
试求 \(E(\xi)\) 和 \(D(\xi)\).
先对 \(\displaystyle\int_{-\infty}^{+\infty} |x|p(x) \mathrm{d}x < +\infty\) 进行验证:
\[
\int_{-\infty}^{+\infty} \frac{|x|}{2\lambda} \mathrm{e}^{-\frac{|x-\mu|}{\lambda}} \mathrm{d}x
\]
由于 \(\dfrac{|x-\mu|}{\lambda} \leqslant 0\) ,且在 \(x\to \infty\) 时趋于 \(-\infty\) ,广义积分收敛.
求其期望即求如下积分:
\[
\begin{aligned}
E(\xi) & = \int_{-\infty}^{+\infty} xp(x) \mathrm{d}x \\
& = \int_{-\infty}^{+\infty} \dfrac{x}{2 \lambda} \mathrm{e}^{-\frac{|x-\mu|}{\lambda}} \mathrm{d}x \\
& = \int_\mu^{+\infty} \dfrac{x}{2 \lambda} \mathrm{e}^{-\frac{x-\mu}{\lambda}} \mathrm{d}x + \int_{-\infty}^\mu \dfrac{x}{2 \lambda} \mathrm{e}^{\frac{x-\mu}{\lambda}}\mathrm{d}x \\
& = \int_{0}^{+\infty} \frac{\lambda}{2} \left(z+\dfrac{\mu}{\lambda}\right)\mathrm{e}^{-z} \mathrm{d}z+ \int_{-\infty}^{0} \frac{\lambda}{2} \left(z+\dfrac{\mu}{\lambda}\right)\mathrm{e}^{z} \mathrm{d}z \\
& = \mu
\end{aligned}
\]
其中最后一步的积分利用分部积分法计算:
\[
\int_{0}^{+\infty} \frac{\lambda}{2} \left(z+\dfrac{\mu}{\lambda}\right)\mathrm{e}^{-z} \mathrm{d}z = -\dfrac{\lambda}{2}\left(z+ \frac{\mu}{\lambda}\right)\mathrm{e}^{-z}\bigg|_0^{+\infty} + \frac{\lambda}{2}\int_0^{+\infty} \mathrm{e}^{-z} \mathrm{d}z = \dfrac{\mu+\lambda}{2}
\]
第二个积分类似可得 \(\dfrac{\mu-\lambda}{2}\) . 因此 \(E(\xi) = \mu\) .
对方差 \(D(\xi)\),考虑
\[
\begin{aligned}
E(\xi^2) & = \int_{-\infty}^{+\infty} \frac{x^2}{2\lambda} \mathrm{e}^{-\frac{|x-\mu|}{\lambda}} \mathrm{d}x \\
& = \int_{0}^{+\infty} (\lambda z+ \mu)^2 \mathrm{e}^{- z} \mathrm{d}z + \int_{-\infty}^{0} (\lambda z+ \mu)^2 \mathrm{e}^{z} \mathrm{d}z \\
& = 2 \lambda^2 + \mu^2
\end{aligned}
\]
那么 \(D(\xi) = 2\lambda^2 +\mu^2 - \mu^2 = 2 \lambda^2\) . \(\square\)
习题四 T12
若 \(a \leqslant \xi \leqslant b\) ,试证:
$$ D(\xi) \leqslant \dfrac{(b-a)^2}{4} $$
并说明等式在何种情况下成立.
定义随机变量 \(\eta\) 为
\[
\eta = \dfrac{\xi-a}{b-a}
\]
从而可知 \(0\leqslant\eta \leqslant 1\) ,且有 \(D(\eta) = \dfrac{1}{(b-a)^2}D(\xi)\) . 那么有
\[
\begin{aligned}
D(\eta) & = E(\eta^2) - [E(\eta)]^2 \\
& \leqslant E(\eta)- [E(\eta)]^2 \\
& \leqslant \frac{1}{4}
\end{aligned}
\]
其中最大值在 \(E(\eta)=\dfrac{1}{2}\) 时取等,那么有
\[
D(\xi) = (b-a)^2 D(\eta) \leqslant \dfrac{(b-a)^2}{4}
\]
根据 \(E(\eta) = \dfrac{1}{b-a}E(\xi)-\dfrac{a}{b-a}\) ,取等条件为
\[
E\left(\xi\right) = \dfrac{a+b}{2}
\]
\(\square\)
习题四 T13
若 \(\xi_1,\xi_2\) 相互独立,均服从 \(N(\mu,\sigma^2)\),试证:
$$ E\max (\xi_1,\xi_2)=\mu+ \dfrac{\sigma}{\sqrt{\pi}} $$
先考虑化为标准正态变量,即
\[
\eta_1 = \frac{\xi_1-\mu}{\sigma},\eta_2 = \frac{\xi_2-\mu}{\sigma}
\]
那么有
\[
E\max (\xi_1,\xi_2) = \sigma E \max (\eta_1,\eta_2) + \mu
\]
那么,\(\max(\eta_1,\eta_2)\) 的密度函数为
\[
2 \varPhi(x)\varphi(x) = \frac{1}{\pi} \mathrm{e}^{- \frac{x^2}{2}} \int_{-\infty}^x \mathrm{e}^{-\frac{u^2}{2}} \mathrm{d}u
\]
故期望可计算为
\[
\begin{aligned}
E \max(\eta_1,\eta_2) & = \int_{-\infty}^{+\infty}\frac{1}{\pi}x \mathrm{e}^{-\frac{x^2}{2}} \left(\int_{-\infty}^{x} \mathrm{e}^{-\frac{u^2}{2}}\mathrm{d}u \right) \mathrm{d}x \\
& = - \frac{1}{\pi}\int_{-\infty}^{+\infty} \left(\int_{-\infty}^{x} \mathrm{e}^{-\frac{u^2}{2}}\mathrm{d}u \right) \mathrm{d} \mathrm{e}^{-\frac{x^2}{2}} \\
& = - \frac{1}{\pi}\left(\int_{-\infty}^{x} \mathrm{e}^{-\frac{u^2}{2}}\mathrm{d}u \right) \mathrm{e}^{-\frac{x^2}{2}} \bigg|_{-\infty}^{+\infty} + \frac{1}{\pi} \int_{-\infty}^{+\infty} \mathrm{e}^{-x^2} \mathrm{d}x \\
& = \frac{1}{\sqrt{\pi}}
\end{aligned}
\]
因此 \(E\max(\xi_1,\xi_2) = \mu+ \dfrac{\sigma}{\sqrt{\pi}}\). \(\square\)
习题四 T14
设 \(f(x)(0\leqslant x < \infty)\) 是单调非降函数,且 \(f(x)>0\). 对随机变量 \(\xi\) ,若 \(E[f(|\xi|)]<\infty\) ,则对任意 \(x>0\) ,
$$ P \left\lbrace |\xi| \geqslant x \right\rbrace \leqslant \dfrac{1}{f(x)} E[f(|\xi|)] $$
习题四 T15
若 \(\xi_1,\xi_2,\cdots,\xi_n\) 为正的独立随机变量,服从相同分布,密度函数为 \(p(x)\),试证:
$$ E\left(\dfrac{\xi_1+\xi_2+\cdots+ \xi_k}{\xi_1+ \xi_2+\cdots+ \xi_n}\right) = \dfrac{k}{n}. $$