跳转至

概率论第八周作业

习题三T26

\(\xi_1\)\(\xi_2\) 是独立随机变量,均服从 Poisson 分布,参数分别为 \(\lambda_1\)\(\lambda_2\) ,尝试直接证明:

(1) \(\xi_1+\xi_2\) 具有 Poisson 分布,参数为 \(\lambda_1+\lambda_2\)

(2) \(P \left\lbrace \xi_1 = k\mid \xi_1+\xi_2 = n \right\rbrace = \displaystyle \binom{n}{k}\left(\dfrac{\lambda_1}{\lambda_1+\lambda_2}\right)^k \left(\dfrac{\lambda_2}{\lambda_1+\lambda_2}\right)^{n-k}\).

(1) 证明
对和的分布,考虑卷积公式,首先对它们的分布密度函数有:

\[ p_i(k) = \dfrac{\lambda_i^k}{k!}\mathrm{e}^{-\lambda_i},i=1,2; k=0,1,2,\cdots \]

\(k<0\) 时,设 \(p_i(k)=0\) ,使用离散的卷积公式有:

\[ \begin{aligned} q(k) & = \sum\limits_{j=-\infty}^{+\infty} p_1(j)p_2(k-j) \\ & = \sum\limits_{j=0}^k p_1(j)p_2(k-j) \\ & = \mathrm{e}^{-(\lambda_1+\lambda_2)}\left(\dfrac{\lambda_1^0}{0!} \dfrac{\lambda_2^k}{k!}+\cdots +\dfrac{\lambda_1^k}{k!}\dfrac{\lambda_2^0}{0!}\right) \\ & = \mathrm{e}^{-(\lambda_1+\lambda_2)}\sum\limits_{j=0}^k \dfrac{\lambda_1^j \lambda_2^{k-j} }{k!}\binom{k}{j} \\ & = \dfrac{(\lambda_1+\lambda_2)^k}{k!}\mathrm{e}^{-(\lambda_1+\lambda_2)} \end{aligned} \]

最后两步利用了二项式定理,由结果可知 \(q(k)\) 为参数为 \(\lambda_1+\lambda_2\) 的 Poisson 分布.

(2) 证明
先考虑求 \(P \left\lbrace \xi_1+\xi_2 = n \right\rbrace\) ,由 (1) 可知取

\[ q(n) = \dfrac{(\lambda_1+\lambda_2)^n}{n!} \mathrm{e}^{-(\lambda_1+\lambda_2)}\tag{26.1} \]

而对 \(P \left\lbrace \xi_1 = k, \xi_2 = n-k \right\rbrace\) ,由二者独立可知:

\[ P \left\lbrace \xi_1 = k, \xi_2 = n-k \right\rbrace = \dfrac{\lambda_1^k}{k!}\cdot \dfrac{\lambda_2^{n-k}}{(n-k)!}\cdot \mathrm{e}^{-(\lambda_1+\lambda_2)} \tag{26.2} \]

那么由条件概率公式有

\[ P \left\lbrace \xi_1 = k\mid \xi_1+\xi_2 = n \right\rbrace = \dfrac{(26.2)}{(26.1)} = \binom{n}{k}\left(\dfrac{\lambda_1}{\lambda_1+\lambda_2}\right)^k \left(\dfrac{\lambda_2}{\lambda_1+\lambda_2}\right)^{n-k} \]

\(\square\)

习题三T28

\(\xi\)\(\eta\) 相互独立且服从同一几何分布,令 \(\zeta = \max (\xi,\eta)\) ,试求:

(1) \((\zeta,\xi)\) 的联合概率分布;

(2) \(\zeta\) 的概率分布;

(3) \(\xi\) 关于 \(\zeta\) 的条件概率分布.


\(\xi\)\(\eta\) 服从的几何分布为:

\[ g(k;p) = (1-p)^{k-1}p, k=1,2,3,\cdots \]

(1) 考虑 \(P \left\lbrace \zeta = x, \xi = y \right\rbrace\)
\(x < y\) ,则 \(P \left\lbrace \zeta = x, \xi = y \right\rbrace = 0\) 是显然的;
\(x = y\) ,那么

\[ \begin{aligned} P \left\lbrace \zeta = x, \xi = y \right\rbrace & = P \left\lbrace \xi = x, \eta \leqslant x \right\rbrace \\ & = P \left\lbrace \xi = x \right\rbrace P \left\lbrace \eta \leqslant x \right\rbrace \\ & = (1-p)^{x-1} p \cdot\sum\limits_{k=1}^x (1-p)^{k-1} p \\ & = p^2 \sum\limits_{k=x-1}^{2x-2} (1-p)^{k} \\ & = p^2 \dfrac{(1-p)^{x-1}[1-(1-p)^x] }{p} \\ & = p (1-p)^{x-1}-p(1-p)^{2x-1} \end{aligned} \]

\(x > y\) ,那么

\[ \begin{aligned} P \left\lbrace \zeta = x, \xi = y \right\rbrace & = P \left\lbrace \xi = y, \eta = x \right\rbrace \\ & = (1-p)^{y-1}p \cdot (1-p)^{x-1}p\\ & = p^2(1-p)^{x+y-2} \end{aligned} \]

因此联合概率分布为:

\[ F_1(x,y) = \begin{cases} 0, & x<y \\ p(1-p)^{x-1}-p(1-p)^{2x-1} , & x=y \\ p^2(1-p)^{x+y-2}, & x> y \end{cases}\quad , x,y\in \mathbb{N^*} \]

(2) \(\zeta\) 的概率分布考虑 \(P \left\lbrace \zeta=x \right\rbrace\) .那么此时有

\[ \begin{aligned} P \left\lbrace \zeta = x \right\rbrace & = P \left\lbrace \xi \leqslant x , \eta =x \right\rbrace + P \left\lbrace \xi = x , \eta \leqslant x \right\rbrace - P \left\lbrace \xi =x , \eta =x \right\rbrace \\ & = 2p(1-p)^{x-1}-2p(1-p)^{2x-1} - p^2(1-p)^{2x-2} \\ \end{aligned} \]

(3) \(\xi\) 关于 \(\zeta\) 的条件概率分布考虑 \(P \left\lbrace \xi = x \mid \zeta = y \right\rbrace\) .

\[ \begin{aligned} P \left\lbrace \xi = x \mid \zeta = y \right\rbrace & = \dfrac{P \left\lbrace \xi = x, \zeta = y \right\rbrace}{P \left\lbrace \zeta = y \right\rbrace} \\ & = \begin{cases} 0, & x> y \\ \dfrac{1-p(1-p)^{y}}{2-2(1-p)^{y} - p(1-p)^{y-1}}, & x=y \\ \dfrac{p(1-p)^{x-1}}{2-2(1-p)^{y} - p(1-p)^{y-1}} , & x< y \end{cases} \end{aligned} \]

\(\square\)

习题三T32

\(\xi,\eta\) 相互独立,分别服从 \(N(0,1)\) ,试证明 \(\psi = \dfrac{\xi}{\eta}\) 服从 Cauchy 分布.

由题可知 \(\xi\) 的分布密度函数为

\[ p_\xi(x) = \dfrac{1}{\sqrt{2 \pi}}\mathrm{e}^{-\frac{x^2}{2}} \]

由于相互独立,可知联合分布密度函数为

\[ p(x,y) = \frac{1}{2\pi} \mathrm{e}^{-\frac{x^2+y^2}{2}} \]

那么根据商的分布有:

\[ \begin{aligned} q(x) & = \int_{-\infty}^{+\infty} |u|\dfrac{1}{2 \pi} \mathrm{e}^{-\frac{u^2x^2+u^2}{2}}\mathrm{d}u \\ & = \frac{1}{\pi}\int_0^{+\infty} u \mathrm{e}^{-\frac{u^2(x^2+1)}{2}}\mathrm{d}u \\ & = \frac{1}{\pi} \dfrac{1}{1+x^2} \end{aligned} \]

因此根据 \(q(x)\) 表达式可知 \(\psi\) 服从 Cauchy 分布. \(\square\)

习题三T38

\(\xi\)\(\eta\) 相互独立,且分别服从 \(\Gamma(r_1,\lambda)\)\(\Gamma(r_2,\lambda)\),试求 \(\alpha = \xi + \eta\)\(\beta = \dfrac{\xi}{\xi+\eta}\) 的联合分布密度函数 \(q(u,v)\) ,并证明:

(1) 随机变量 \(\beta\) 服从 \(\beta\) 分布:$$ p(v) = \dfrac{\Gamma(r_1+r_2)}{\Gamma(r_1)\Gamma(r_2)}v^{r_1-1}(1-v)^{r_2-1}, \qquad 0<v<1 $$.

(2) 随机变量 \(\alpha\)\(\beta\) 独立.

解:
首先解得

\[ \begin{cases} \xi = \alpha \beta \\ \eta = \alpha(1-\beta) \end{cases} \]

故 Jacobi 行列式为:

\[ J = \begin{vmatrix} \beta & \alpha \\ 1-\beta & -\alpha \end{vmatrix} = -\alpha \]

另一方面,\((\xi,\eta)\) 分别服从 \(\Gamma(r_1,\lambda),\Gamma(r_2,\lambda)\) 且相互独立,则联合分布密度为:

\[ p(x_1,x_2) = \dfrac{\lambda^{r_1}}{\Gamma(r_1)}x_1^{r_1-1}\mathrm{e}^{-\lambda x_1} \cdot \dfrac{\lambda^{r_2}}{\Gamma(r_2)}x_2^{r_2-1}\mathrm{e}^{-\lambda x_2} \]

\[ \begin{aligned} q(u,v) & = |J| p (uv, u(1-v)) \\ & = u \dfrac{\lambda^{r_1}}{\Gamma(r_1)}(uv)^{r_1-1}\mathrm{e}^{-\lambda uv} \cdot \dfrac{\lambda^{^{r_2}}}{\Gamma(r_2)} [u(1-v)]^{r_2-1}\mathrm{e}^{-\lambda u(1-v)} \\ & = \dfrac{\lambda^{r_1+r_2}}{\Gamma(r_1)\Gamma(r_2)} \mathrm{e}^{- \lambda u}u^{r_1+r_2-1}v^{r_1-1}(1-v)^{r_2-1} \end{aligned}\tag{38.1} \]

由课堂结论:

\((\xi,\eta)\) 的联合密度为 \(f(x,y)\) ,且 \(f(x,y)=f_1(x)f_2(y)\) ,则 \(\xi,\eta\) 独立.

\((38.1)\) 可以分离有:

\[ q(u,v) = \dfrac{\lambda^{r_1+r_2}}{\Gamma(r_1+r_2)}u^{r_1+r_2-1}\mathrm{e}^{-\lambda u}\cdot \dfrac{\Gamma(r_1+r_2)}{\Gamma(r_1)\Gamma(r_2)}v^{r_1-1}(1-v)^{r_2-1}\tag{38.2} \]

\(\alpha\)\(\beta\) 相互独立.

下面考虑先求 \(\alpha\) 的分布,利用卷积公式有:

\[ \begin{aligned} p_\alpha(y) & = \int_{-\infty}^{+\infty}p(x,y-x)\mathrm{d}x \\ & = \int_{-\infty}^{+\infty} \dfrac{\lambda^{r_1+r_2}}{\Gamma(r_1)\Gamma(r_2)}\mathrm{e}^{-\lambda y}x^{r_1-1}(y-x)^{r_2-1}\mathrm{d}x \\ & = \dfrac{\lambda^{r_1+r_2}}{\Gamma(r_1)\Gamma(r_2)}\mathrm{e}^{-\lambda y} y^{r_1+r_2-1}\int_{0}^{1} \left(\frac{x}{y}\right)^{r_1-1}\left(1-\dfrac{x}{y}\right)^{r_2-1}\mathrm{d}\dfrac{x}{y} \\ & = \dfrac{\lambda^{r_1+r_2}}{\Gamma(r_1)\Gamma(r_2)}\mathrm{e}^{-\lambda y} y^{r_1+r_2-1} \dfrac{\Gamma(r_1)\Gamma(r_2)}{\Gamma(r_1+r_2)} \\ & = \dfrac{\lambda^{r_1+r_2}}{\Gamma(r_1+r_2)}\mathrm{e}^{-\lambda y} y^{r_1+r_2-1} \end{aligned} \]

根据 \((38.2)\) 可知 \(\beta\) 服从 \(\beta\) 分布. \(\square\)

习题三T39

\(\xi,\eta\) 独立,且均服从 \(N(0,1)\) ,试求 \(U = \xi^2+\eta^2\)\(V = \dfrac{\xi}{\eta}\) 的密度函数,并证明它们是独立的.

先求 Jacobi 矩阵的逆:

\[ \boldsymbol{J}^{-1}=\begin{pmatrix} \dfrac{\partial U}{\partial \xi} & \dfrac{\partial U}{\partial \eta} \\ \dfrac{\partial V}{\partial \xi} & \dfrac{\partial V}{\partial \eta}\end{pmatrix} = \begin{pmatrix}2 \xi & 2 \eta \\ \dfrac{1}{\eta} & -\dfrac{\xi}{\eta^2}\end{pmatrix} \]

那么行列式为:

\[ |J| = |\boldsymbol{J}^{-1}|^{-1} = \frac{1}{2}\frac{\eta^2}{\xi^2+\eta^2} = \frac{1}{2}\dfrac{1}{1+V^2} \]

\(\xi\)\(\eta\) 独立,且均服从 \(N(0,1)\) ,那么联合密度为:

\[ p(x,y) = \dfrac{1}{\sqrt{2 \pi}}\mathrm{e}^{-\frac{x^2}{2}} \dfrac{1}{\sqrt{2 \pi}}\mathrm{e}^{-\frac{y^2}{2}} = \frac{1}{2\pi}\mathrm{e}^{-\frac{x^2+y^2}{2}} \]

因此只需解出 \(\xi^2\)\(\eta^2\) 即可:

\[ \begin{cases} \xi^2 = U- \dfrac{U}{1+V^2} \\ \eta^2 = \dfrac{U}{1+V^2} \end{cases} \]

因此 \(U\)\(V\) 的联合密度为:

\[ \begin{aligned} q(u,v) & = 2\cdot\dfrac{1}{2}\dfrac{1}{1+v^2}\cdot \dfrac{1}{2\pi}\mathrm{e}^{-\frac{u}{2}} \end{aligned} \]

其中的 \(2\)\(\eta>0\)\(\eta<0\) 的两种情形带来的,表达式已分离,故相互独立,根据

\[ \int_{-\infty}^{+\infty} \dfrac{1}{1+v^2}\mathrm{d}v = \pi \]

\[ \int_{0}^{+\infty} \mathrm{e}^{- \frac{u}{2}}\mathrm{d}u = 2 \]

有密度函数

\[ p_U (x) = \frac{1}{2}\mathrm{e}^{-\frac{u}{2}},p_V(y) = \frac{1}{\pi}\frac{1}{1+y^2} \]

\(\square\)

习题三T43

\(\xi\)\(\eta\) 是相互独立相同分布的随机变量,其密度函数不等于 \(0\) 且有二阶导数,试证明若 \(\xi+\eta\)\(\xi-\eta\) 相互独立,则随机变量 \(\xi,\eta,\xi+\eta,\xi-\eta\) 均服从正态分布.

\[ \begin{cases} u = \xi + \eta \\ v = \xi- \eta \end{cases} \]

那么有

\[ \begin{cases} \xi = \dfrac{1}{2}u + \dfrac{1}{2}v \\ \eta = \dfrac{1}{2}u - \dfrac{1}{2}v \end{cases} \]

从而 Jacobi 行列式为 \(J = -\dfrac{1}{2}\) .

\(\xi\) 的分布密度函数为 \(p(x)\) ,那么 \(\xi\)\(\eta\) 的联合分布密度记为 \(p(x)p(y)\) ,此时,设 \(\xi+\eta\) 的分布密度函数为 \(q(x)\)\(\xi-\eta\) 的分布密度函数为 \(r(x)\) ,那么有

\[ q(u)r(v) = \frac{1}{2} p\left(\frac{1}{2}u + \frac{1}{2}v\right) p\left(\frac{1}{2}u - \frac{1}{2}v\right) \]

取对数分离

\[ \ln q(u)+ \ln r(v) = -\ln 2 + \ln p\left(\frac{1}{2}u + \frac{1}{2}v\right) + \ln p\left(\frac{1}{2}u - \frac{1}{2}v\right) \]

\(h(z) = \ln p(z)\) ,此时对 \(u\) 求偏导有

\[ \dfrac{q'(u)}{q(u)} = \frac{1}{2}h'\left(\frac{1}{2}u + \frac{1}{2}v\right)+ \frac{1}{2}h' \left(\frac{1}{2}u - \frac{1}{2}v\right) \]

再对 \(v\) 求偏导:

\[ h'' \left(\frac{1}{2}u - \frac{1}{2}v\right) + h'' \left(\frac{1}{2}u - \frac{1}{2}v\right) = 0 \]

\(u=v\)

\[ h''(u) = h''(0) \]

此时记 \(h''(0)=\lambda\) ,那么有

\[ h(u) = \dfrac{\lambda}{2}x^2 + C_1 x +C_2 \]

\[ p(u) = \mathrm{e}^{\frac{\lambda}{2}x^2 + C_1 x +C_2} = \mathrm{e}^{C_2 + \frac{C_1^2}{2\lambda}} \mathrm{e}^{\frac{\lambda}{2}(x-\frac{C_1}{\lambda})^2} \]

因此 \(p(u)\) 满足正态分布的分布密度函数形式,从而 \(\xi\)\(\eta\) 均服从正态分布. 而对 \(\xi+\eta\)\(\xi-\eta\) ,它们均为 \(\xi\)\(\eta\) 的线性组合,从而为正态分布. \(\square\)

习题三T45

设随机变量 \(\xi\)\(\eta\) 相互独立,\(\xi\sim B(n,p)\) ,而 \(\eta\) 服从 \((0,1)\) 上均匀分布,试求 \(\xi+\eta\) 的分布函数和密度函数.

易知

\[ P \left\lbrace \xi = k \right\rbrace = \binom{n}{k} p^k (1-p)^{n-k} \]

此时考虑 \(\xi+\eta\) 的分布函数,下面分类讨论:对 \(P \left\lbrace \xi+\eta <x \right\rbrace\)

\(x\in (-\infty,0)\) 时, \(P \left\lbrace \xi+\eta<x \right\rbrace = 0\) 是显然的.

\(x\in (0,n+1)\) 时,将 \(x\) 分解为 \(x = [x]+\left\lbrace x \right\rbrace\) ,其中 \([x]\) 为 Gauss 取整,\(\left\lbrace x \right\rbrace\)\(x\) 的小数部分. 那么

\[ \begin{aligned} P \left\lbrace \xi+ \eta < x \right\rbrace & = P \left\lbrace \xi = [x], \eta< \left\lbrace x \right\rbrace \right\rbrace + P \left\lbrace \xi < [x], \eta \leqslant 1 \right\rbrace \\ & = \left\lbrace x \right\rbrace\cdot \binom{n}{[x]}p^{[x]} (1-p)^{n-[x]} +\sum\limits_{j=0}^{[x]-1} \binom{n}{j} p^j (1-p)^j \end{aligned} \]

\(x\in [n+1,+\infty)\) 时,\(P \left\lbrace \xi+\eta<x \right\rbrace = 1\) .

因此分布函数为:

\[ F(x) = \begin{cases} 0, & x \leqslant 0 \\ \displaystyle\left\lbrace x \right\rbrace\cdot \binom{n}{[x]}p^{[x]} (1-p)^{n-[x]} +\sum\limits_{j=0}^{[x]-1} \binom{n}{j} p^j (1-p)^j , & 0< x < n+1 \\ 1, & x \geqslant n+1 \end{cases} \]

对分布密度函数,根据分布函数可得

\[ p(x) = \begin{cases} 0, & x \leqslant 0, x> n+1 \\ \displaystyle\binom{n}{[x]} p^{[x]}(1-p)^{n-[x]}, & 0< x \leqslant n+1 \end{cases} \]

\(\square\)

习题三 T46

试求顺序统计量 \(\xi_k^*\)\(\xi_l^*(k<l)\) 的联合密度函数.

\(x<y\) ,考虑化为多项分布有

\[ P \left\lbrace x \leqslant \xi_k^* < x+ \Delta x, y \leqslant \xi_l^* < y + \Delta y \right\rbrace \]

那么比 \(\xi_k^*\) 小的 \(k-1\) 个变量小于 \(x\) ,在 \(\xi_k^*\)\(\xi_l^*\) 之间的 \(l-k-1\) 个变量在 \(x+ \Delta x\)\(y\) 之间,比 \(\xi_l^*\) 大的 \(n-l\) 个变量大于 \(y+\Delta y\) . 这两个对应的算式也是简单的:

\[ \begin{aligned} P \left\lbrace x \leqslant \xi_k^* < x+ \Delta x, y \leqslant \xi_l^* < y + \Delta y \right\rbrace & = \dfrac{n!}{(k-1)!(l-k-1)!(n-l)!} \\ & \cdot [F(x)]^{k-1} [F(x+\Delta x)-F(x)] \\ & \cdot [F(y)-F(x+\Delta x)]^{l-k-1} \\ & \cdot [F(y+\Delta y)-F(y)] \\ & \cdot [1- F(y+\Delta y)]^{n-l} \end{aligned} \]

其中,含有 \(\Delta x\) 的项考虑

\[ F(x+ \Delta x)-F(x) = p(x)\Delta x \]

\[ p_{(\xi_k^*, \xi_l^*)}(x,y) \Delta x \Delta y = P \left\lbrace x \leqslant \xi_k^* < x+ \Delta x, y \leqslant \xi_l^* < y + \Delta y \right\rbrace \cdot \]

代入有

\[ p_{(\xi_k^*, \xi_l^*)}(x,y) = \begin{cases} \dfrac{n!}{(k-1)!(l-k-1)!(n-l)!} [F(x)]^{k-1} \cdot \\ [F(y)-F(x)]^{l-k-1} \cdot [1-F(y)]^{n-l} p(x)p(y), & x<y \\ 0, & x \geqslant y \end{cases} \]

\(\square\)

习题四 T2

随机变量 \(\mu\) 取非负整数值 \(n \geqslant 0\) 的概率为 \(p_n = A\dfrac{B^n}{n!}\) ,已知 \(E(\mu) = a\) ,试决定 \(A\)\(B\) .

根据期望的定义可列出如下两式:

\[ \begin{cases} \sum\limits_{k=0}^\infty p_k =\sum\limits_{k=0}^\infty A\dfrac{B^k}{k!} =1 \\ \sum\limits_{k=1}^\infty kp_k = AB\sum\limits_{k=1}^\infty \dfrac{B^{k-1}}{(k-1)!} = a \end{cases} \]

根据其中的第二个式子可得 \(B = a\) ,代入第一个式子可得

\[ A \sum\limits_{k=0}^\infty \dfrac{a^k}{k!} = A \mathrm{e}^a =1 \]

\(A = \mathrm{e}^{-a}\) . \(\square\)

习题四 T5

一袋中含有 \(a\) 只白球,\(b\) 只黑球,从中取出 \(c\) 只 (\(c \leqslant a+b\)) ,求摸出白球数 \(\mu\) 的数学期望.

首先求

\[ P( \mu = k) = \dfrac{\displaystyle\binom{a}{k}\binom{b}{c-k}}{\displaystyle\binom{a+b}{c}} \]

根据离散型随机变量的数学期望定义,有

\[ \begin{aligned} E(\mu) & = \dfrac{1}{\displaystyle\binom{a+b}{c}}\sum\limits_{k=1}^{\min(a,c)} k \binom{a}{k}\binom{b}{c-k} \\ & = \dfrac{1}{\displaystyle \binom{a+b}{c}}\sum\limits_{k=1}^{\min(a,c)} a \binom{a-1}{k-1}\binom{b}{c-k} \\ & = \dfrac{a}{\displaystyle \binom{a+b}{c}} \sum\limits_{k=1}^{\min(a,c)}\binom{a-1}{k-1}\binom{b}{c-k} \end{aligned} \]

利用 Vande Monde 恒等式,可得

\[ E(\mu) = \dfrac{a}{\displaystyle\binom{a+b}{c}}\binom{a+b-1}{c-1} = \dfrac{ac}{a+b} \]

故数学期望为 \(\dfrac{ac}{a+b}\) . \(\square\)

习题四 T7

试证:若取非负整数值的随机变量 \(\xi\) 的数学期望存在,则
$$ E(\xi) = \sum\limits_{k=1}^\infty P \left\lbrace \xi \geqslant k \right\rbrace. $$

首先根据定义有

\[ \begin{aligned} E(\xi) & = \sum\limits_{k=1}^\infty k\cdot P \left\lbrace \xi = k \right\rbrace \\ & = \lim_{n\to \infty} \sum\limits_{k=1}^n k \cdot P \left\lbrace \xi = k \right\rbrace \\ & = \lim_{n\to \infty} \sum\limits_{k=1}^n k \cdot \left[P \left\lbrace \xi \geqslant k \right\rbrace - P \left\lbrace \xi \geqslant k+1 \right\rbrace\right] \end{aligned} \]

Abel 求和公式,可得

\[ E(\xi) = \lim_{n\to \infty} \sum\limits_{k=1}^n P \left\lbrace \xi \geqslant k \right\rbrace = \sum\limits_{k=1}^\infty P \left\lbrace \xi \geqslant k \right\rbrace \]

\(\square\)

评论