中国科学技术大学-432统计学-2021年

一、(15分) 甲、乙两人掷骰子, 甲先掷, 如果扔到1则把骰子给乙, 否则自己继续, 同样乙也扔到1就给甲, 问第nn次是甲投掷的概率.

Solution: 用事件 AiA_{i} 表示第 ii 次是甲投郑, 则已知 P(A1)=1P\left(A_{1}\right)=1 以及 P(A2)=56P\left(A_{2}\right)=\frac{5}{6}. 根据全概率公式, 有

P(An)=P(AnAn1)P(An1)+P(AnAn1)P(An1)=56P(An1)+16(1P(An1))=23P(An1)+16\begin{aligned} P\left(A_{n}\right) &=P\left(A_{n} \mid A_{n-1}\right) P\left(A_{n-1}\right)+P\left(A_{n} \mid \overline{A_{n-1}}\right) P\left(\overline{A_{n-1}}\right) \\ &=\frac{5}{6} P\left(A_{n-1}\right)+\frac{1}{6}\left(1-P\left(A_{n-1}\right)\right) \\ &=\frac{2}{3} P\left(A_{n-1}\right)+\frac{1}{6} \end{aligned}

解此差分方程得 P(An)=(23)n112+12P\left(A_{n}\right)=\left(\frac{2}{3}\right)^{n-1} \cdot \frac{1}{2}+\frac{1}{2}.

二、(15分) 已知XU(0,1),YU(β,1),Z1U(0,β),Z2U(0,β)X\sim U(0,1),Y\sim U(\beta,1),Z_1\sim U(0,\beta),Z_2\sim U(0,\beta)且它们相互独立, 记Vi=I{Xβ}Zi+I{X>β}YV_i=I_{\{X\le \beta\}}Z_i+I_{\{X>\beta\}}Y, i=1,2i=1,2.

(1) 证明ViU(0,1),i=1,2V_i\sim U(0,1),i=1,2;

(2) 求Corr(V1,V2)(V_1,V_2).

Solution: (1) 已知 V1=I{Xβ}Z1+I{X>β}YV_{1}=I_{\{X \leqslant \beta\}} Z_{1}+I_{\{X>\beta\}} Y, 且 X,Y,Z1,Z2X, Y, Z_{1}, Z_{2} 相互独立, 则

P(V1v)=P(V1v,Xβ)+P(V1v,X>β)=P(Z1v)P(Xβ)+P(Yv)P(X>β)\begin{aligned} P\left(V_{1} \leqslant v\right) &=P\left(V_{1} \leqslant v, X \leqslant \beta\right)+P\left(V_{1} \leqslant v, X>\beta\right) \\ &=P\left(Z_{1} \leqslant v\right) P(X \leqslant \beta)+P(Y \leqslant v) P(X>\beta) \end{aligned}

若 (i) 0<vβ,P(V1v)=vββ+0=v0<v \leqslant \beta, P\left(V_{1} \leqslant v\right)=\frac{v}{\beta} \cdot \beta+0=v;
(ii) β<v1,P(V1<v)=β+(1β)vβ1β=v\beta<v \leqslant 1, P\left(V_{1}<v\right)=\beta+(1-\beta) \cdot \frac{v-\beta}{1-\beta}=v;
(iii) v>1,P(V1<v)=1v>1, P\left(V_{1}<v\right)=1.
FV1(v)={0,v0v,0<v11,v>1F_{V_{1}}(v)=\left\{\begin{array}{ll}0, & v \leqslant 0 \\ v, & 0<v \leqslant 1 \\ 1, & v>1\end{array}\right.V1U(0,1)V_{1} \sim U(0,1), 同理 V2U(0,1)V_{2} \sim U(0,1).

(2) E(V1)=E(V2)=12,Var(V1)=Var(V2)=112E\left(V_{1}\right)=E\left(V_{2}\right)=\frac{1}{2}, \operatorname{Var}\left(V_{1}\right)=\operatorname{Var}\left(V_{2}\right)=\frac{1}{12}.
根据重期望公式,

E(V1V2)=E(E(V1V2X))=E(V1V2X>β)P(X>β)+E(V1V2Xβ)P(Xβ)=E(Y2)(1β)+E(Z1Z2)β=[(1+β2)2+(1β)212](1β)+(β2)2β=4β312Corr(V1,V2)=Cov(V1,V2)Var(V1)Var(V2)=E(V1V2)E(V1)E(V2)1/12=1β3\begin{aligned} E\left(V_{1} V_{2}\right) &=E\left(E\left(V_{1} V_{2} \mid X\right)\right) \\ &=E\left(V_{1} V_{2} \mid X>\beta\right) P(X>\beta)+E\left(V_{1} V_{2} \mid X \leqslant \beta\right) P(X \leqslant \beta) \\ &=E\left(Y^{2}\right)(1-\beta)+E\left(Z_{1} Z_{2}\right) \beta \\ &=\left[\left(\frac{1+\beta}{2}\right)^{2}+\frac{(1-\beta)^{2}}{12}\right](1-\beta)+\left(\frac{\beta}{2}\right)^{2} \beta \\ &=\frac{4-\beta^{3}}{12} \\ \operatorname{Corr}\left(V_{1}, V_{2}\right) &=\frac{\operatorname{Cov}\left(V_{1}, V_{2}\right)}{\sqrt{\operatorname{Var}\left(V_{1}\right)} \cdot \sqrt{\operatorname{Var}\left(V_{2}\right)}}=\frac{E\left(V_{1} V_{2}\right)-E\left(V_{1}\right) E\left(V_{2}\right)}{1 / 12} \\ &=1-\beta^{3} \end{aligned}

三、(15分) 有线性模型Yi=β0+β1Xi+εiY_i=\beta_0 + \beta_1 X_i +\varepsilon_i, 其中εi\varepsilon_i独立同服从N(0,σ2)N(0,\sigma^2), 考虑β0,β1,σ2\beta_0,\beta_1,\sigma^2的最小二乘估计, 回答下述问题:

(1) 所有自变量XiX_i都增加2, 最小二乘估计将会怎么变化?

(2) 所有自变量XiX_i都乘2, 最小二乘估计将会怎么变化?

Solution:
(1) (β0,β1)\left(\beta_{0}, \beta_{1}\right) 的最小二乘估计是

{β^0=yˉxˉβ^1β^1=i=1n(xixˉ)(yiyˉ)i=1n(xixˉ)2\left\{\begin{array}{l} \hat{\beta}_{0}=\bar{y}-\bar{x} \hat{\beta}_{1} \\ \hat{\beta}_{1}=\frac{\sum_{i=1}^{n}\left(x_{i}-\bar{x}\right)\left(y_{i}-\bar{y}\right)}{\sum_{i=1}^{n}\left(x_{i}-\bar{x}\right)^{2}} \end{array}\right.

xix_{i} 增加 2, 可以发现 β^1\hat{\beta}_{1} 不改变; 而 β^0=yˉ(xˉ+2)β^1=β^02β^1\hat{\beta}_{0}^{*}=\bar{y}-(\bar{x}+2) \hat{\beta}_{1}=\hat{\beta}_{0}-2 \hat{\beta}_{1}.
(2) 当 xix_{i} 变为原来的 2 倍时, 则

β^1=i=1n(2xi2xˉ)(yiyˉ)i=1n(2xi2xˉ)2=12β^1\hat{\beta}_{1}^{*}=\frac{\sum_{i=1}^{n}\left(2 x_{i}-2 \bar{x}\right)\left(y_{i}-\bar{y}\right)}{\sum_{i=1}^{n}\left(2 x_{i}-2 \bar{x}\right)^{2}}=\frac{1}{2} \hat{\beta}_{1}

β^0=yˉ(2xˉ)β^1=yˉxˉβ^1=β^0\hat{\beta}_{0}^{*}=\bar{y}-(2 \bar{x}) \hat{\beta}_{1}^{*}=\bar{y}-\bar{x} \hat{\beta}_{1}=\hat{\beta}_{0}.

四、(15分) 已知X,YN(0,σ2)X,Y\sim N(0,\sigma^2)且相互独立, 令θ=arcsinXX2+Y2\theta=\arcsin \frac{X}{\sqrt{X^2+Y^2}}, 试

(1) 求θ\theta的概率密度函数;

(2) 判断θ\thetaX2+Y2X^2+Y^2是否独立.

Solution: [提示] 该题需要注意反正切函数的值域.

(1) 作变换 {r=X2+Y2θ=arcsin(XX2+Y2){X=rsinθY=rcosθ{X=rsinθY=rcosθ\left\{ \begin{array}{l} r=\sqrt{X^2+Y^2}\\ \theta =\arcsin \left( \frac{X}{\sqrt{X^2+Y^2}} \right)\\ \end{array}\Longleftrightarrow \left\{ \begin{array}{l} X=r\sin \theta\\ Y=r\cos \theta\\ \end{array} \right. \text{或}\left\{ \begin{array}{l} X=r\sin \theta\\ Y=-r\cos \theta\\ \end{array} \right. \right., 其中 r>0,π2<θ<π2r > 0, -\frac{\pi}{2}< \theta < \frac{\pi}{2}. 两种情况的雅各比行列式均为 J=rJ = r.

(X,Y)(X, Y) 的联合密度函数是 fX,Y(x,y)=12πσ2ex2+y2σ2,<x,y<+f_{X, Y}(x, y)=\frac{1}{2 \pi \sigma^{2}} e^{-\frac{x^{2}+y}{2 \sigma^{2}}},-\infty<x, y<+\infty, 因此根 据公式法 (R,θ)(R, \theta) 的联合密度是

fR,θ(r,θ)=rfX,Y(rsinθ,rcosθ)+rfX,Y(rsinθ,rcosθ)=1πσ2er22σ2r,r>0,π2<θ<π2.\begin{aligned} f_{R,\theta}\left( r,\theta \right) &=rf_{X,Y}\left( r\sin \theta ,r\cos \theta \right) +rf_{X,Y}\left( -r\sin \theta ,-r\cos \theta \right) \\ &=\frac{1}{\pi \sigma ^2}e^{-\frac{r^2}{2\sigma ^2}}r,r>0,-\frac{\pi}{2}<\theta <\frac{\pi}{2}. \end{aligned}

rr 积分, 得到 θ\theta 的边际密度函数

fθ(θ)=0+fR,θ(r,θ)dr=1π,π2<θ<π2.f_{\theta}\left( \theta \right) =\int_0^{+\infty}{f}_{R,\theta}\left( r,\theta \right) dr=\frac{1}{\pi},-\frac{\pi}{2}<\theta <\frac{\pi}{2}.

(2) 由于 fR,θ(r,θ)=fR(r)fθ(θ)f_{R, \theta}(r, \theta)=f_{R}(r) f_{\theta}(\theta), 所以 R,θR, \theta 相互独立, 因此有 R2,θR^{2}, \theta 相互独立, 即 X2+Y2X^{2}+Y^{2}θ\theta 相互独立.

五、(15分) 某电子元件的寿命服从期望为λ\lambda的指数分布, 从这批原件中抽取nn件作寿命试验, 规定到第r(0<rn)r(0<r\le n)个电子元件失效时就停止试验, 这样获取了前rr个次序统计量X(1),,X(r)X_{(1)},\cdots,X_{(r)}, 求

(1) λ\lambda的最大似然估计λ^\hat{\lambda};

(2) 若r=2r=2, 判断λ^\hat{\lambda}是否为λ\lambda的无偏估计.

Solution: (1) 对应事件是 {X(1)=t1,X(2)=t2,,X(r)=tr}\left\{X_{(1)}=t_{1}, X_{(2)}=t_{2}, \cdots, X_{(r)}=t_{r}\right\}, 因此似然函数可写为

L(λ)=f1,,r(t1,,trλ)L(\lambda)=f_{1, \cdots, r}\left(t_{1}, \cdots, t_{r} \mid \lambda\right)

其中 f1,,r(t1,,trλ)f_{1, \cdots, r}\left(t_{1}, \cdots, t_{r} \mid \lambda\right)(X(1),,X(r))\left(X_{(1)}, \cdots, X_{(r)}\right) 的联合密度函数, 它是

f1,,r(t1,,trλ)=n!(nr)!1λrei=1rtiλ(etrλ)nr,0<t1<<tr=n!(nr)!1λre1λ(i=1ti+(nr)tr),0<t1<<tr\begin{aligned} f_{1, \cdots, r}\left(t_{1}, \cdots, t_{r} \mid \lambda\right) &=\frac{n !}{(n-r) !} \frac{1}{\lambda^{r}} e^{-\frac{\sum_{i=1}^{r} t_{i}}{\lambda}}\left(e^{-\frac{t_{r}}{\lambda}}\right)^{n-r}, 0<t_{1}<\cdots<t_{r} \\ &=\frac{n !}{(n-r) !} \frac{1}{\lambda^{r}} e^{-\frac{1}{\lambda}\left(\sum_{i=1}^{t_{i}}+(n-r) t_{r}\right)}, 0<t_{1}<\cdots<t_{r} \end{aligned}

lnL(λ)=ln(n!(nr)!)rlnλ1λ(i=1rti+(nr)tr)\ln L(\lambda)=\ln \left(\frac{n !}{(n-r) !}\right)-r \ln \lambda-\frac{1}{\lambda}\left(\sum_{i=1}^{r} t_{i}+(n-r) t_{r}\right), 令

dlnL(λ)dλ=rλ+i=1rti+(nr)trλ2=0\frac{d \ln L(\lambda)}{d \lambda}=-\frac{r}{\lambda}+\frac{\sum_{i=1}^{r} t_{i}+(n-r) t_{r}}{\lambda^{2}}=0

解得 λ\lambda 的极大似然估计是 λ^=i=1rX(i)+(nr)X(r)r\hat{\lambda}=\frac{\sum_{i=1}^{r} X_{(i)}+(n-r) X_{(r)}}{r}.

(2) 当 r=2r=2 时, λ^=X(1)+(n1)X(2)2\hat{\lambda}=\frac{X_{(1)}+(n-1) X_{(2)}}{2}, 而

nX(1)Exp(1λ),(n1)(X(2)X(1))Exp(1λ)n X_{(1)} \sim \operatorname{Exp}\left(\frac{1}{\lambda}\right),(n-1)\left(X_{(2)}-X_{(1)}\right) \sim \operatorname{Exp}\left(\frac{1}{\lambda}\right)

所以 EX(1)=λn,E[(n1)X(2)]=λ+n1nλE X_{(1)}=\frac{\lambda}{n}, E\left[(n-1) X_{(2)}\right]=\lambda+\frac{n-1}{n} \lambda, 则

Eλ^=EX(1)+E[(n1)X(2)]2=λE \hat{\lambda}=\frac{E X_{(1)}+E\left[(n-1) X_{(2)}\right]}{2}=\lambda

λ^\hat{\lambda}λ\lambda 的无偏估计.

六、(15分) 甲化肥是某化肥厂研制的畅销款, 现该化肥厂研制出改进品种乙化肥, 为探究乙化肥 的效果是否好于甲化肥,你的小组被委任展开对比实验, 第一组选取 13 块稻 田, 施用甲化肥,其亩产平均值为 12.3,12.3, 样本方差为 $10 ; $ 第二组选取 11 块稻田, 施用乙化肥, 其 亩产平均为 14.4,14.4, 样本方差为 7.8.7.8 . 假设亩产服从正态分布. 在显著性水平选择 α=0.1\alpha=0.1的情况下, 问:

(1) 是否能够认为两组的方差相等?

(2) 是否能够认为乙化肥的均值高于甲化肥?

Solution:
(1) 用随机变量 X1,,X13X_{1}, \cdots, X_{13} 表示施用甲化肥稻田的亩产, 用随机变量 Y1,,Y11Y_{1}, \cdots, Y_{11} 表示施用乙 化肥稻田的亩产.
假设 XiX_{i} i.i.d N(μ1,σ12),Yi\sim N\left(\mu_{1}, \sigma_{1}^{2}\right), Y_{i} i.i.d N(μ2,σ22)\sim N\left(\mu_{2}, \sigma_{2}^{2}\right), 考虑假设检验问题

H0:σ12=σ22H1:σ12σ22H_{0}: \sigma_{1}^{2}=\sigma_{2}{ }^{2} \longleftrightarrow H_{1}: \sigma_{1}{ }^{2} \neq \sigma_{2}{ }^{2}

检验统计量是 F=s12s22=107.8=1.28F=\frac{s_{1}^{2}}{s_{2}^{2}}=\frac{10}{7.8}=1.28, 检验的拒绝域是

W={FFα2(m1,n1)}{FF1α2(m1,n1)}={FF0.05(12,10)}{FF0.95(12,10)}={F2.91}{F12.75}\begin{aligned} W &=\left\{F \geqslant F_{\frac{\alpha}{2}}(m-1, n-1)\right\} \cup\left\{F \leqslant F_{1-\frac{\alpha}{2}}(m-1, n-1)\right\} \\ &=\left\{F \geqslant F_{0.05}(12,10)\right\} \cup\left\{F \leqslant F_{0.95}(12,10)\right\} \\ &=\{F \geqslant 2.91\} \cup\left\{F \leqslant \frac{1}{2.75}\right\} \end{aligned}

所以 FWF \notin W, 因此我们不能拒绝原假设, 应认为两者方差相等.
(2) 考虑假设检验问题 H0:μ1μ20H1:μ1μ2<0H_{0}: \mu_{1} -\mu_2 \ge0 \longleftrightarrow H_{1}: \mu_{1}-\mu_{2}<0. 检验统计量为

T=XˉYˉsw1m+1n=12.314.4(m1)s12+(n1)s22m+n2113+111=12.314.41210+107.813+112113+111=1.709\begin{aligned} T &=\frac{\bar{X}-\bar{Y}}{s_{w} \sqrt{\frac{1}{m}+\frac{1}{n}}} \\ &=\frac{12.3-14.4}{\sqrt{\frac{(m-1) s_{1}^{2}+(n-1) s_{2}^{2}}{m+n-2} \sqrt{\frac{1}{13}+\frac{1}{11}}}} \\ &=\frac{12.3-14.4}{\sqrt{\frac{12 \cdot 10+10 \cdot 7.8}{13+11-2}} \cdot \sqrt{\frac{1}{13}+\frac{1}{11}}}=-1.709 \end{aligned}

检验的拒绝域为

W={Ttα(m+n2)}={Tt0.1(m+n2)}={T1.3212}W=\left\{T \leqslant-t_{\alpha}(m+n-2)\right\}=\left\{T \leqslant-t_{0.1}(m+n-2)\right\}=\{T \leqslant-1.3212\}

所以 TWT \in W, 我们应拒绝原假设, 可以认为乙化肥带来的亩产均值高于甲化肥.

七、(20分) 二维随机变量 (X,Y)(X,Y) 的密度函数为 f(x,y)=Ae(3x+4y)x>0,y>0f(x, y)=A e^{-(3x+4 y)} x>0, y>0.

(1) 求系数 AA;

(2) 问XXYY 是否独立?

(3) 试求$ Z=X+Y $的密度函数;

(4) 试求 E(XX+Y=1)E(X|X+Y=1).

Solution:
(1) 根据概率密度函数的正则性有 1=0+dx0+Ae(3x+4y)dy=A121=\int_{0}^{+\infty} d x \int_{0}^{+\infty} A e^{-(3 x+4 y)} d y= \frac{A}{12}, 解得 A=12A=12.

(2) (X,Y)(X, Y) 的联合密度函数 f(x,y)=3e3x4e4y,x,y>0f(x, y)=3 e^{-3 x} 4 e^{-4 y}, x, y>0 是变量可分离的, 即有 f(x,y)=fX(x)fY(y)f(x, y)=f_{X}(x) f_{Y}(y), 所以它们相互独立.

(3) 考虑变量变换 {Z=X+YU=X{X=UY=ZU\left\{\begin{array}{l}Z=X+Y \\ U=X\end{array} \Leftrightarrow\left\{\begin{array}{l}X=U \\ Y=Z-U\end{array}\right.\right., 变换的雅各比行列式为 J=1011=1J=\left|\begin{array}{cc}1 & 0 \\ -1 & 1\end{array}\right|=1, 因此由公式法, (Z,U)(Z, U) 的联合密度函数是

fZ,U(z,u)=f(u,zu)1,u>0,zu>0=12e(4zu),z>u>0\begin{aligned} f_{Z, U}(z, u) &=f(u, z-u)|1|, u>0, z-u>0 \\ &=12 e^{-(4 z-u)}, z>u>0 \end{aligned}

因此 ZZ 的边际密度函数是

fZ(z)=0z12e(4zu)du=12e4z(ez1),z>0f_{Z}(z)=\int_{0}^{z} 12 e^{-(4 z-u)} d u=12 e^{-4 z}\left(e^{z}-1\right), z>0

(4) E(XX+Y=1)=E(UZ=1)E(X \mid X+Y=1)=E(U \mid Z=1), 而 UUZ=1Z=1 条件下的条件密度函数是

fUZ=1(u)=fU,Z(u,1)fZ(1)=12e(4u)12e4(e1)=eue1,0<u<1f_{U \mid Z=1}(u)=\frac{f_{U, Z}(u, 1)}{f_{Z}(1)}=\frac{12 e^{-(4-u)}}{12 e^{-4}(e-1)}=\frac{e^{u}}{e-1}, 0<u<1

所以 E(UZ=1)=01ueue1du=1e1E(U \mid Z=1)=\int_{0}^{1} u \frac{e^{u}}{e-1} d u=\frac{1}{e-1}.

八、(20分) 甲、乙、丙三个工厂生产同一种产品, 产品质量分别为三个等级(1,2,3分别代表高, 中, 低). 今从三个厂中共抽 300 件产品, 逐件检测, 得结果如下:

 级数  甲工厂  乙工厂  丙工厂  总和 1 级 3638301042 级 4033351083 级 18264488 总和 9497109300\begin{array}{|c|c|c|c|c|} \hline \text { 级数 } & \text { 甲工厂 } & \text { 乙工厂 } & \text { 丙工厂 } & \text { 总和 } \\ \hline 1 \text { 级 } & 36 & 38 & 30 & 104 \\ \hline 2 \text { 级 } & 40 & 33 & 35 & 108 \\ \hline 3 \text { 级 } & 18 & 26 & 44 & 88 \\ \hline \text { 总和 } & 94 & 97 & 109 & 300 \\ \hline \end{array}

(1) 三个厂的产品质量是否一致?

(2) 若不一致, 问哪个厂的质量更优, 哪个厂更劣?

Solution:
(1) 考虑假设检验问题 H0H_{0} : 三个工厂产品质量一致.
这实际上意味着: 产品质量与工厂独立. 检验统计量 χ2=n[i=13j=13nij2ni.n.j1]=12.275\chi^{2}=n\left[\sum_{i=1}^{3} \sum_{j=1}^{3} \frac{n_{i j}^{2}}{n_{i} . n_{. j}}-1\right]=12.275, 检验的拒绝域是 W={χ2χα2((r1)(s1))}={χ2χ0.052(4)}={χ29.488}W=\left\{\chi^{2} \geqslant \chi_{\alpha}^{2}((r-1)(s-1))\right\}=\left\{\chi^{2} \geqslant \chi_{0.05}^{2}(4)\right\}=\left\{\chi^{2} \geqslant 9.488\right\} 于是 χ2W\chi^{2} \in W, 故可以拒绝原假设, 即认为三个工厂产品质量不一致.
(2) 考虑用均值作为衡量指标, 则

Xˉ甲 =36×1+40×2+18×394=1.80851Xˉ乙 =38×1+33×2+26×394=1.93617Xˉ丙 =30×1+35×2+44×3109=2.12844\begin{aligned} &\bar{X}_{\text {甲 }}=\frac{36 \times 1+40 \times 2+18 \times 3}{94}=1.80851 \\ &\bar{X}_{\text {乙 }}=\frac{38 \times 1+33 \times 2+26 \times 3}{94}=1.93617 \\ &\bar{X}_{\text {丙 }}=\frac{30 \times 1+35 \times 2+44 \times 3}{109}=2.12844 \end{aligned}

愈小者愈优, 故甲最优, 丙最劣.

九、(20分) 有来自总体N(0,σ2)N(0,\sigma^2)的随机样本X1,,XnX_1,\cdots,X_n, 其次序统计量是X(1),,X(n)X_{(1)},\cdots,X_{(n)}是其次序统计量, RiR_iXiX_i的秩, 即Ri=jR_i=j当且仅当Xi=X(j)X_i=X_{(j)}. 回答下述问题.

(1) 求E(Ri)E(R_i);

(2) 求Cov(Ri,Rj)Cov(R_i,R_j);

(3) 判断(X(1),,X(n))(X_{(1)},\cdots,X_{(n)})(R1,,Rn)(R_1,\cdots,R_n)是否独立.

Solution: (1) 根据对称性, XiX_{i}nn 个样本中排在任意一个位置的概率都一致, 即

P(Xi=X(j))=1n,j=1,2,,nP\left(X_{i}=X_{(j)}\right)=\frac{1}{n}, j=1,2, \ldots, n

所以有 P(Ri=j)=1n,j=1,2,,nP\left(R_{i}=j\right)=\frac{1}{n}, j=1,2, \cdots, n, 因此 E(Ri)=j=1n1n=n+12E\left(R_{i}\right)=\sum_{j=1}^{n} \frac{1}{n}=\frac{n+1}{2}.

(2) 当 j=ij=i 时, 有

Cov(Ri,Ri)=Var(Ri)=E(Ri2)(ERi)2=(n+1)(2n+1)6(n+1)24=n2112\begin{aligned} \operatorname{Cov}\left(R_{i}, R_{i}\right)&=\operatorname{Var}\left(R_{i}\right) \\ &=E\left(R_{i}^{2}\right)-\left(E R_{i}\right)^{2}=\frac{(n+1)(2 n+1)}{6}-\frac{(n+1)^{2}}{4}=\frac{n^{2}-1}{12} \end{aligned}

iji \neq j 时, 根据对称性有

P(Ri=m,Rj=k)=P(Ri=1,Rj=2)=1n1n1,mkP\left(R_{i}=m, R_{j}=k\right)=P\left(R_{i}=1, R_{j}=2\right)=\frac{1}{n} \cdot \frac{1}{n-1}, m \neq k

E(RiRj)=1n(n1)[m=1nk=1nmkm=1nm2]=112(n+1)(3n+2)E\left(R_{i} R_{j}\right)=\frac{1}{n(n-1)}\left[\sum_{m=1}^{n} \sum_{k=1}^{n} m k-\sum_{m=1}^{n} m^{2}\right]=\frac{1}{12}(n+1)(3 n+2). 于是

Cov(Ri,Rj)=E(RiRj)E(Ri)E(Rj)=112(n+1)\operatorname{Cov}\left(R_{i}, R_{j}\right)=E\left(R_{i} R_{j}\right)-E\left(R_{i}\right) E\left(R_{j}\right)=-\frac{1}{12}(n+1) \text {. }

(3) (R1,,Rn)\left(R_{1}, \cdots, R_{n}\right) 的联合分布列是, 根据对称性有

P(R1=i1,R2=i2,,Rn=in)=P(R1=1,R2=2,,Rn=n)=1n!P\left(R_{1}=i_{1}, R_{2}=i_{2}, \cdots, R_{n}=i_{n}\right)=P\left(R_{1}=1, R_{2}=2, \cdots, R_{n}=n\right)=\frac{1}{n !}

其中 {i1,i2,,in}\left\{i_{1}, i_{2}, \cdots, i_{n}\right\}{1,2,,n}\{1,2, \cdots, n\} 的任意一种排列.
假设总体密度函数是 f(x)f(x), 则 (X(1),X(2),,X(n),R1,R2,,Rn)\left(X_{(1)}, X_{(2)}, \cdots, X_{(n)}, R_{1}, R_{2}, \cdots, R_{n}\right) 的联合密度函数是

h(x1,,xn,R1=i1,,Rn=in)=f(X1,X2,,Xn)(xi1,xi2,,xin),x1<<xn=f(xi1)f(xin),x1<<xn=i=1nf(xi),x1<<xn\begin{aligned} h\left( x_1,\cdots ,x_n,R_1=i_1,\cdots ,R_n=i_n \right) &=f_{\left( X_1,X_2,\cdots ,X_n \right)}\left( x_{i_1},x_{i_2},\cdots ,x_{i_n} \right) ,x_1<\cdots <x_n\\ &=f\left( x_{i_1} \right) \cdots f\left( x_{i_n} \right) ,x_1<\cdots <x_n\\ &=\prod_{i=1}^n{f}\left( x_i \right) ,x_1<\cdots <x_n\\ \end{aligned}

(X(1),,X(n))\left(X_{(1)}, \cdots, X_{(n)}\right) 的联合密度函数是我们熟知的,

f(X(1),X(n))(x1,,xn)=n!i=1nf(xi),x1<<xnf_{\left(X_{(1)}, \cdots X_{(n)}\right)}\left(x_{1}, \cdots, x_{n}\right)=n ! \prod_{i=1}^{n} f\left(x_{i}\right), x_{1}<\cdots<x_{n}

因此

h(x1,,xn,R1=i1,,Rn=in)=f(X(1),X(n))(x1,,xn)P(R1=i1,,Rn=in)h\left(x_{1}, \cdots, x_{n}, R_{1}=i_{1}, \cdots, R_{n}=i_{n}\right)=f_{\left(X_{(1)}, \cdots X_{(n)}\right)}\left(x_{1}, \cdots, x_{n}\right) \cdot P\left(R_{1}=i_{1}, \cdots, R_{n}=i_{n}\right)

所以 (X(1),,X(n))\left(X_{(1)}, \cdots, X_{(n)}\right)(R1,,Rn)\left(R_{1}, \cdots, R_{n}\right) 是相互独立的.