数理统计与数据分析第三版习题 第3章 第43-53题

题目43

令$U_1$和$U_2$是$[0,1]$上相互独立的均匀随机变量计算并画出$S=U_1+U_2$的密度函数

解题思路

$$f_x(x)=\{\begin{array}{r}10\le x\le 1\\ 0其它\end{array}$$

$$f_y(y)=\{\begin{array}{r}10\le y\le 1\\ 0其它\end{array}$$
卷积公式
$$f_z(z)={\int }_{-\infty }^{+\infty }f(u,z-u)du={\int }_{-\infty }^{+\infty }{f}_x(u)f_y(z-u)du$$
计算时只要计算被积函数非零的部分
即$0\le u\le 1且0\le z-u\le 1$ 可以以写成$0\le u\le 1且z-1\le u\le z$
当$0\le z\le 1$时 $z-1$最小可以为-1但根据$0\le u$这个条件所以$u$的下限取0，上限取z
当$1\le z\le 2$时 $z-1$最小可以0满足u必须大于0的要求，所以此时$u$的下限为$z-1$而上限$z$最大可以取2不满足$u\le 1$的要求，取上限为1

$$f_z(z)={\int }_{-\infty }^{+\infty }{f}_x(u)f_y(z-u)du=\{\begin{array}{rl}& {\int }_0^{z}1du=z,0\le z\le 1\\ & {\int }_{z-1}^{1}1du=2-z,1\le z\le 2\\ & 0,其它\end{array}$$

题目44

假设$X$和$Y$是独立的离散的随机变量，并且取值0，1，2时的概率都是$\frac{1}{3}$,计算$X+Y$的频率函数

解题思路

根据卷积定义
$$p_Z(z)=\sum _{x=-\infty }^{\infty }{p}_X(x)p_Y(z-x)$$
本题主要是要确定x求合的范围
根据题意$Z=X+Y$,则$0\le z\le 4$并且要满足$0\le x\le 2$
当$z\le 2,0\le x\le z$
当$z\ge 2,z-2\le x\le 2$
所以,当$z\le 2$
$$\begin{array}{rl}{p}_Z(z)& =\sum _{x=0}^z{p}_X(x)p_Y(z-x)\\ & =\sum _{x=0}^z\frac{1}{3}\cdot \frac{1}{3}\\ & =(z+1)\cdot \frac{1}{9}\\ & =\frac{1}{9}z+\frac{1}{9}\end{array}$$

当$z\ge 2$
$$\begin{array}{rl}{p}_Z(z)& =\sum _{x=z-2}^2{p}_X(x)p_Y(z-x)\\ & =\sum _{x=z-2}^2\frac{1}{3}\cdot \frac{1}{3}\\ 上限减去下限再加1就是总的相加的次数& \\ & =(2-(z-2)+1)\cdot \frac{1}{9}\\ & =\frac{1}{9}(5-z)\end{array}$$

题目45

对于泊松分布，假设事件用$A$和$B$独立的标识出来，且满足概率$p_A+p_B=1$.如果泊松分布的参数是$\lambda$.证明:标识为$A$事件数据服从参数为$p_A\lambda$的泊松分布

解题思路

根据题意，假设做了100次实验发生了15次，在这15次中事件是A 是7次事件B是8次。而不是事件发生记作A，事件不发生记作B

且满足概率$p_A+p_B=1$，则表示发生若干次事件中，只有A和B不会再有其他的类型事件了

$$\begin{array}{rl}& N=A+B\sim Poission(\lambda )\\ 在N发生的情况下，& A事件只是按固定概率发生，A事件发生的次数服务二项分布\\ & P(A|N)\sim Bin(N,p_A)\\ p(A=a)& =\sum _0^{\infty }P(A=a|N=n)\\ & =\sum _0^{\infty }\left(\genfrac{}{}{0ex}{}{n}{a}\right){p_A}^a(1-p_a{)}^{n-a}\cdot \frac{e^{-\lambda }\cdot {\lambda }^n}{n!}\\ & =\sum _0^{\infty }\frac{n!}{a!\cdot (n-a)!}\cdot {p_A}^a(1-p_a{)}^{n-a}\cdot \frac{e^{-\lambda }\cdot {\lambda }^n}{n!}\\ & =\sum _0^{\infty }\frac{{p_A}^a(1-P_A{)}^{n-a}\cdot e^{-\lambda }\cdot {\lambda }^n}{a!\cdot (n-a)!}\\ 把{\lambda }^n& 分成{\lambda }^{n-a}\cdot {\lambda }^a\\ & =\sum _0^{\infty }\frac{{p_A}^a(1-P_A{)}^{n-a}\cdot e^{-\lambda }\cdot {\lambda }^{n-a}\cdot {\lambda }^a}{a!\cdot (n-a)!}\\ & =\sum _0^{\infty }\frac{{p_A}^a\cdot e^{-\lambda }\cdot {\lambda }^a}{a!}\cdot \frac{(1-p_A{)}^{n-a}\cdot {\lambda }^{n-a}}{(n-a)!}\cdot \frac{e^{-(1-p_A)\lambda }}{e^{-(1-p_A)\lambda }}\\ & =\underset{前}{\underbrace{\frac{{p_A}^a\cdot e^{-\lambda }\cdot {\lambda }^a}{a!\cdot e^{-(1-p_A)\lambda }}\cdot }}\underset{后}{\underbrace{\sum _0^{\infty }\frac{(1-p_A{)}^{n-a}\cdot {\lambda }^{n-a}\cdot e^{-(1-p_A)\lambda }}{(n-a)!\cdot }}}\\ 本式& 的后半部分简化：令t=n-a\\ & \sum _{t=0}^{\infty }\frac{e^{-(1-p_A)\lambda }\cdot [(1-p_A)\lambda {]}^t}{t!}=1\\ 本式& 前半部分简化后(后半式为1，所以前半部分也是最终结果)：\\ & \frac{(\lambda p_A{)}^a\cdot e^{-p_A\lambda }}{a!}\\ 所以& 本题结果是服务参数为p_A\lambda 的泊松分布\end{array}$$

题目46

令$T_1$和$T_2$分别是参数为${\lambda }_1$和${\lambda }_2$的独立指数分布，计算$T_1+T_2$的密度函数

解题思路

根据卷积公式
$$f_z(z)={\int }_{-\infty }^{+\infty }{f}_x(x)f_y(z-x)dx$$
接下来确定x的积分范围，根据指数分布前提x>0 然后 z-x>0,即 x<z,所以x的积分范围是从0到z
$$\begin{array}{rl}{f}_z(z)& ={\int }_0^z{f}_x(x)f_y(z-x)dx\\ & ={\int }_0^z{\lambda }_1{e}^{-{\lambda }_{1}x}\cdot {\lambda }_2{e}^{-{\lambda }_2(z-x)}dx\\ & ={\lambda }_1{\lambda }_2{\int }_0^z{e}^{-{\lambda }_{1}x}\cdot e^{-{\lambda }_2(z-x)}dx\\ & ={\lambda }_1{\lambda }_2{\int }_0^z{e}^{-{\lambda }_{1}x-{\lambda }_2(z-x)}dx\\ & ={\lambda }_1{\lambda }_2{\int }_0^z{e}^{-{\lambda }_{1}x-{\lambda }_{2}z+{\lambda }_{2}x}dx\\ & ={\lambda }_1{\lambda }_2{e}^{-{\lambda }_{2}z}{\int }_0^z{e}^{{\lambda }_{2}x-{\lambda }_{1}x}dx\\ & ={\lambda }_1{\lambda }_2{e}^{-{\lambda }_{2}z}{\int }_0^z{e}^{({\lambda }_2-{\lambda }_1)x}dx\\ & ={\lambda }_1{\lambda }_2{e}^{-{\lambda }_{2}z}\cdot \frac{1}{{\lambda }_2-{\lambda }_1}{\int }_0^z{e}^{({\lambda }_2-{\lambda }_1)x}d(({\lambda }_2-{\lambda }_1)x)\\ & =\frac{{\lambda }_1{\lambda }_2}{{\lambda }_2-{\lambda }_1}\cdot e^{-{\lambda }_{2}z}\cdot e^{({\lambda }_2-{\lambda }_1)x}{|}_0^z\\ & =\frac{{\lambda }_1{\lambda }_2}{{\lambda }_2-{\lambda }_1}{e}^{-{\lambda }_{2}z}(e^{({\lambda }_2-{\lambda }_1)z}-1)\\ & =\frac{{\lambda }_1{\lambda }_2}{{\lambda }_2-{\lambda }_1}(e^{-{\lambda }_{1}z}-e^{-{\lambda }_{2}z})\end{array}$$

题目47

令$X$和$Y$是独立的标准的正态随机变量，计算$Z=X+Y$的密度，并证明:$Z$ 也是正态分布(提示在积分过程 中利用配方技术)

解题思路

利用卷积公式
$$\begin{array}{rl}{f}_z(z)& ={\int }_{-\infty }^{+\infty }{f}_x(x)f_y(z-x)dx\\ & ={\int }_{-\infty }^{+\infty }\frac{1}{\sqrt{2\pi }}{e}^{-\frac{x^2}{2}}\cdot \frac{1}{\sqrt{2\pi }}{e}^{-\frac{(z-x{)}^2}{2}}\\ & =\frac{1}{2\pi }{\int }_{-\infty }^{+\infty }{e}^{-\frac{x^2}{2}-\frac{(z-x{)}^2}{2}}dx\\ & =\frac{1}{2\pi }{\int }_{-\infty }^{+\infty }{e}^{(-x^2-\frac{1}{2}{z}^2+xz)}dx\\ & =\frac{1}{2\pi }{\int }_{-\infty }^{+\infty }{e}^{-(x-\frac{z}{2}{)}^2-\frac{1}{4}{z}^2}dx\\ & =\frac{1}{2\pi }{e}^{\frac{z^2}{2}}{\int }_{-\infty }^{+\infty }{e}^{-(x-\frac{z}{2}{)}^2}dx\\ 令u=(x-\frac{z}{2})则：& \\ & =\frac{1}{2\pi }{e}^{\frac{z^2}{2}}{\int }_{-\infty }^{+\infty }{e}^{-u^2}du\\ {\int }_{-\infty }^{+\infty }{e}^{-u^2}du与z没有关系所以算出来是一个常数c& \\ & =c\frac{1}{2\pi }{e}^{\frac{z^2}{2}}\end{array}$$
所以本题结果是一个均值为0标准差为1的正态分布

题目48

令$N_1$和$N_2$是独立随机变量，分别服从参数为${\lambda }_1$和${\lambda }_2$，证明：$N=N_1+N_2$的分布是参数为${\lambda }_1+{\lambda }_2$的泊松分布

解题思路

利用卷积公式
根据题意确定上下限，$x\ge 0且z-x\ge 0$,所以$x\le z$
$$\begin{array}{rl}{p}_Z(z)& =\sum _{x=-\infty }^{\infty }{p}_X(x)p_Y(z-x)\\ & =\sum _{x=0}^z\frac{{\lambda }_1^x{e}^{-{\lambda }_1}}{x!}\cdot \frac{{\lambda }_2^{z-x}{e}^{-{\lambda }_2}}{(z-x)!}\\ & =\sum _{x=0}^z\frac{e^{-{\lambda }_1-{\lambda }_2}{\lambda }_1^x{\lambda }_2^{z-x}}{x!(z-x)!}\\ & =e^{-{\lambda }_1-{\lambda }_2}\sum _{x=0}^z\frac{1}{x!(z-x)!}\cdot {\lambda }_1^x{\lambda }_2^{z-x}\\ & =e^{-{\lambda }_1-{\lambda }_2}\sum _{x=0}^z\frac{\left(\genfrac{}{}{0ex}{}{z}{x}\right)}{z!}\cdot {\lambda }_1^x{\lambda }_2^{z-x}\\ & =\frac{e^{-{\lambda }_1-{\lambda }_2}}{z!}\sum _{x=0}^z\left(\genfrac{}{}{0ex}{}{z}{x}\right)\cdot {\lambda }_1^x{\lambda }_2^{z-x}\\ & =\frac{e^{-{\lambda }_1-{\lambda }_2}}{z!}\cdot ({\lambda }_1+{\lambda }_2{)}^z\\ & =\frac{e^{-({\lambda }_1+{\lambda }_2)}({\lambda }_1+{\lambda }_2{)}^z}{z!}\end{array}$$

题目49

计算$Z=X+Y$的密度函数，其中 $X和Y$的联合密度由3.3节中的3.3.4给出

解题思路

找到书中对应的密度函数
$$f(x,y)=\{\begin{array}{rl}& {\lambda }^2{e}^{-\lambda y}0\le x\le y,\lambda >0\\ & 0其它情况\end{array}$$
分析条件y=z-x 且 x<y 则 $x<z-x=>x<\frac{z}{2}$

$$\begin{array}{rl}{f}_Z(z)& ={\int }_0^{\frac{z}{2}}f(x,z-x)dx\\ & ={\int }_0^{\frac{z}{2}}{\lambda }^2{e}^{-\lambda (z-x)}dx\\ & ={\lambda }^2\cdot \frac{1}{-\lambda \cdot -1}{\int }_0^{\frac{z}{2}}{e}^{-\lambda (z-x)}d(-\lambda (z-x))\\ & =\lambda \cdot (e^{-\lambda \frac{z}{2}}-e^{-\lambda z})\end{array}$$

题目50

令X和Y是且有联合分布的连续随机变量，求Z=X-Y的密度表达式

解题思路

思路1：
$$\begin{array}{rl}{F}_Z(z)& =P(Z\le z)=P(X-Y\le z)\\ & =\underset{x-y\le z}{\iint }f(x,y)dxdy\\ & ={\int }_{-\infty }^{\infty }{\int }_{-\infty }^{y+z}f(x,y)dxdy\\ & ={\int }_{-\infty }^{\infty }\left[{\int }_{-\infty }^{y+z}f(x,y)dx\right]dy\\ 令u& =x-y\\ & ={\int }_{-\infty }^{\infty }\left[{\int }_{-\infty }^{z}f(y+u,y)du\right]dy\\ f_Z(z)& =F_Z^{{}^{\prime }}(z)={\int }_{-\infty }^{\infty }f(y+z,y)dy\end{array}$$
参考：https://wenku.baidu.com/view/d86049500c22590102029df1.html
思路2：
$$\begin{array}{rl}{F}_Z(z)& =P(Z\le z)=P(X-Y\le z)\\ & =\underset{x-y\le z}{\iint }f(x,y)dxdy\\ & ={\int }_{-\infty }^{\infty }{\int }_{x-z}^{\infty }f(x,y)dydx\\ & ={\int }_{-\infty }^{\infty }\left[{\int }_{x-z}^{\infty }f(x,y)dy\right]dx\\ 令u=x-y& \\ & ={\int }_{-\infty }^{\infty }\left[{\int }_z^{-\infty }f(x,x-u)du\cdot -1\right]dx\\ & ={\int }_{-\infty }^{\infty }\left[{\int }_{-\infty }^{z}f(x,x-u)du\right]dx\\ f_Z(z)& =F_Z^{{}^{\prime }}(z)={\int }_{-\infty }^{\infty }f(x,x-z)dx\end{array}$$

题目51

令$X$和$Y$是且有联密度函数$f(x,y),Z=XY$证明:Z的密度函数为
$$f_Z(z)={\int }_{-\infty }^{\infty }f(y,\frac{z}{y})\frac{1}{|y|}dy$$

解题思路

$$\begin{array}{rl}P(Z\le z)& =P(XY\le z)\\ & =P(X\le \frac{z}{y}|Y>0)+P(X>\frac{z}{y}|Y<0)\\ & ={\int }_0^{\infty }dy{\int }_{-\infty }^{\frac{z}{y}}f(x,y)dx+{\int }_{-\infty }^{0}dy{\int }_{\frac{z}{y}}^{\infty }f(x,y)dx\\ 因此:& \\ f_Z(z)& ={\int }_0^{\infty }dy\cdot \frac{1}{y}f(\frac{z}{y},y)-{\int }_{-\infty }^{0}dy\cdot \frac{1}{y}f(\frac{z}{y},y)\\ 如果\underset{y\to 0}{\lim }\frac{f(\frac{z}{y},y)}{y}有限，& 则上式可以简写成\\ f_Z(z)& ={\int }_{-\infty }^{\infty }f(\frac{z}{y},y)\frac{1}{|y|}dy\end{array}$$

题目52

计算两个独立随机变量商的密度

解题思路

假定两个变量是标准的均匀密度即在【0，1】之间

根据书中公式
$$f_Z(z)={\int }_{-\infty }^{\infty }|x|f_X(x)f_Y(xz)dx$$
标准密度即$f_Z(x)=1,f_Y(y)=1$,根据题意$0\le x\le 1$
$$\begin{array}{rl}{f}_Z(z)& ={\int }_0^{1}x\cdot 1\cdot 1dx\\ & =\frac{1}{2}\end{array}$$

题目53

考虑用两种方法构建两个矩形，令$U_1,U_2,U_3$是$[0,1]$上独立的均匀随机变量，一个矩形的连长为$U_1和U_2$，另一个是连长为$U_3$的正方形，计算正方形面积大于矩形面积的概率

解题思路

根据题意：$P(U_3^2>U_2{U}_3)=P(U_2>\sqrt{U_2{U}_1})$
$$\begin{array}{rl}P(U_2>\sqrt{U_2{U}_1})& ={\int }_0^1{\int }_0^1{\int }_{\sqrt{u_2{u}_1}}^{1}f(u_1,u_2,u_3)d{u}_{3}d{u}_{2}d{u}_1\\ & ={\int }_0^1{\int }_0^1{\int }_{\sqrt{u_2{u}_1}}^{1}d{u}_{3}d{u}_{2}d{u}_1\\ & ={\int }_0^1{\int }_0^{1}1-{\sqrt{u}}_2\cdot {\sqrt{u}}_{1}d{u}_{2}d{u}_1\\ & ={\int }_0^1{\left[u_2-\frac{2\sqrt{u_1}{u_2}^{\frac{3}{2}}}{3}\right]}_0^{1}d{u}_1\\ & ={\int }_0^{1}1-\frac{2\sqrt{u_1}}{3}d{u}_1\\ & ={\left[u_1-\frac{2}{3}\cdot \frac{2{u_1}^{\frac{3}{2}}}{3}\right]}_0^1\\ & =\frac{5}{9}\end{array}$$