一、定义与性质

设 X 为 随 机 变 量 , I 是 一 个 包 含 0 的 ( 有 限 或 无 限 的 ) 开 区 间 , 对 任 意 t ∈ I , 期 望 E e t x 存 在 设X为随机变量,I是一个包含0的(有限或无限的)开区间,对任意t∈I,期望Ee^{tx}存在 XI0()tIEetx
则 称 函 数 M X ( t ) = E ( e t X ) = ∫ − ∞ + ∞ e t x d F ( x ) , t ∈ I 为 X 的 矩 母 函 数 则称函数M_{X}(t)=E(e^{tX})=\int_{-\infin}^{+\infin}e^{tx}dF(x),t∈I为X的矩母函数 MX(t)=E(etX)=+etxdF(x),tIX
设 X 为 任 意 随 机 变 量 , 称 函 数 φ X ( t ) = E ( e i t X ) = ∫ − ∞ + ∞ e i t x d F ( x ) 为 X 的 特 征 函 数 设X为任意随机变量,称函数\varphi_{X}(t)=E(e^{itX})=\int_{-\infin}^{+\infin}e^{itx}dF(x)为X的特征函数 XφX(t)=E(eitX)=+eitxdF(x)X
一 个 随 机 变 量 的 矩 母 函 数 不 一 定 存 在 , 但 是 特 征 函 数 一 定 存 在 。 一个随机变量的矩母函数不一定存在,但是特征函数一定存在。
随 机 变 量 与 特 征 函 数 存 在 一 一 对 应 的 关 系 随机变量与特征函数存在一一对应的关系

二、离散型随机变量的分布

0、退化分布(Degenerate distribution)

若 X 服 从 参 数 为 a 的 退 化 分 布 , 那 么 f ( k ; a ) = { 1 , k = a 0 , k ≠ a 若X服从参数为a的退化分布,那么f(k;a)=\left\{\begin{matrix} 1,k=a \\ 0,k\neq a \end{matrix}\right. Xa退f(k;a)={1,k=a0,k=a
M ( t ) = e t a M(t)=e^{ta} M(t)=eta
φ ( t ) = e i t a \varphi(t)=e^{ita} φ(t)=eita
M ′ ( t ) = a e t a M'(t)=ae^{ta} M(t)=aeta
E X = M ′ ( 0 ) = a EX=M'(0)=a EX=M(0)=a
M ′ ′ ( t ) = a 2 e t a M''(t)=a^2e^{ta} M(t)=a2eta
E X 2 = M ′ ′ ( 0 ) = a 2 EX^2=M''(0)=a^2 EX2=M(0)=a2
D X = E X 2 − ( E X ) 2 = 0 DX=EX^2-(EX)^2=0 DX=EX2(EX)2=0

1、离散型均匀分布(Discrete uniform distribution)

若 X 服 从 离 散 型 均 匀 分 布 D U ( a , b ) , 则 X 分 布 函 数 为 F ( k ; a , b ) = ⌊ k ⌋ − a + 1 b − a + 1 若X服从离散型均匀分布DU(a,b) ,则X分布函数为F(k;a,b)=\frac{\lfloor k\rfloor -a+1}{b-a+1} XDU(a,b),XF(k;a,b)=ba+1ka+1
则 矩 母 函 数 M ( t ) = ∑ k = a b e t k P ( x = k ) 则矩母函数M(t)=\sum_{k=a}^{b} e^{tk}P(x=k) M(t)=k=abetkP(x=k)
= ( ∑ k = a b e t k ) 1 b − a + 1 =(\sum_{k=a}^{b} e^{tk})\frac{1}{b-a+1} =(k=abetk)ba+11
= e a t − e ( b + 1 ) t ( 1 − e t ) ( b − a + 1 ) =\frac{e^{at}-e^{(b+1)t}}{(1-e^{t})(b-a+1)} =(1et)(ba+1)eate(b+1)t
特 征 函 数 φ ( t ) = ∑ k = a b e i t k P ( x = k ) 特征函数\varphi(t)=\sum_{k=a}^{b} e^{itk}P(x=k) φ(t)=k=abeitkP(x=k)
= ( ∑ k = a b e i t k ) 1 b − a + 1 =(\sum_{k=a}^{b} e^{itk})\frac{1}{b-a+1} =(k=abeitk)ba+11
= e a i t − e ( b + 1 ) i t ( 1 − e i t ) ( b − a + 1 ) =\frac{e^{ait}-e^{(b+1)it}}{(1-e^{it})(b-a+1)} =(1eit)(ba+1)eaite(b+1)it
M ′ ( t ) = 1 b − a + 1 ( a e a t − ( b + 1 ) e ( b + 1 ) t ) ( 1 − e t ) + ( e a t − e ( b + 1 ) t ) e t ( e t − 1 ) 2 M'(t)=\frac{1}{b-a+1}\frac{(ae^{at}-(b+1)e^{(b+1)t})(1-e^t)+(e^{at}-e^{(b+1)t})e^t}{(e^{t}-1)^{2}} M(t)=ba+11(et1)2(aeat(b+1)e(b+1)t)(1et)+(eate(b+1)t)et
t = 0 为 M ′ ( t ) 的 可 去 间 断 点 , 补 充 定 义 M ′ ( 0 ) = lim ⁡ t → 0 M ′ ( t ) t=0为M'(t)的可去间断点,补充定义M'(0)=\lim_{t\rightarrow0}M'(t) t=0M(t)M(0)=t0limM(t)
E X = M ′ ( 0 ) = lim ⁡ t → 0 1 b − a + 1 ( a 2 e a t − ( b + 1 ) 2 e ( b + 1 ) t ) ( 1 − e t ) + ( e a t − e ( b + 1 ) t ) e t 2 ( e t − 1 ) e t EX=M'(0)=\lim_{t\rightarrow0}\frac{1}{b-a+1}\frac{(a^2e^{at}-(b+1)^2e^{(b+1)t})(1-e^t)+(e^{at}-e^{(b+1)t})e^t}{2(e^{t}-1)e^t} EX=M(0)=t0limba+112(et1)et(a2eat(b+1)2e(b+1)t)(1et)+(eate(b+1)t)et
= lim ⁡ t → 0 1 b − a + 1 ( a 2 e a t − ( b + 1 ) 2 e ( b + 1 ) t ) ( e − t − 1 ) + ( e a t − e ( b + 1 ) t ) 2 ( e t − 1 ) =\lim_{t\rightarrow0}\frac{1}{b-a+1}\frac{(a^2e^{at}-(b+1)^2e^{(b+1)t})(e^{-t}-1)+(e^{at}-e^{(b+1)t})}{2(e^{t}-1)} =t0limba+112(et1)(a2eat(b+1)2e(b+1)t)(et1)+(eate(b+1)t)
= lim ⁡ t → 0 1 b − a + 1 ( a 3 e a t − ( b + 1 ) 3 e ( b + 1 ) t ) ( e − t − 1 ) − ( a 2 e a t − ( b + 1 ) 2 e ( b + 1 ) t ) e − t + ( a e a t − ( b + 1 ) e ( b + 1 ) t ) 2 e t =\lim_{t\rightarrow0}\frac{1}{b-a+1}\frac{(a^3e^{at}-(b+1)^3e^{(b+1)t})(e^{-t}-1)-(a^2e^{at}-(b+1)^2e^{(b+1)t})e^{-t}+(ae^{at}-(b+1)e^{(b+1)t})}{2e^{t}} =t0limba+112et(a3eat(b+1)3e(b+1)t)(et1)(a2eat(b+1)2e(b+1)t)et+(aeat(b+1)e(b+1)t)
= − a 2 + ( b + 1 ) 2 + a − ( b + 1 ) 2 ( b − a + 1 ) =\frac{-a^2+(b+1)^2+a-(b+1)}{2(b-a+1)} =2(ba+1)a2+(b+1)2+a(b+1)
= − a 2 + ( b + 1 ) 2 2 ( b − a + 1 ) − 1 2 =\frac{-a^2+(b+1)^2}{2(b-a+1)}-\frac{1}{2} =2(ba+1)a2+(b+1)221
= ( b + 1 − a ) ( b + 1 + a ) 2 ( b − a + 1 ) − 1 2 =\frac{(b+1-a)(b+1+a)}{2(b-a+1)}-\frac{1}{2} =2(ba+1)(b+1a)(b+1+a)21
= b + 1 + a 2 − 1 2 =\frac{b+1+a}{2}-\frac{1}{2} =2b+1+a21
= b + a 2 =\frac{b+a}{2} =2b+a
由 于 对 M ′ ( t ) 求 导 得 到 M ′ ′ ( t ) , 再 求 M ′ ′ ( 0 ) 的 方 法 比 较 繁 琐 , 而 我 们 只 需 要 t = 0 时 M 的 二 阶 导 数 值 , 由于对M'(t)求导得到M''(t),再求M''(0)的方法比较繁琐,而我们只需要t=0时M的二阶导数值, M(t)M(t)M(0)t=0M
因 此 可 以 考 虑 使 用 T a y l o r 公 式 计 算 M ′ ′ ( 0 ) 因此可以考虑使用Taylor公式计算M''(0) 使TaylorM(0)
令 1 − e t = u , t = 0 时 , u = 0 令1-e^t=u,t=0时,u=0 1et=u,t=0,u=0
M ( t ) = e a t − e ( b + 1 ) t ( 1 − e t ) ( b − a + 1 ) M(t)=\frac{e^{at}-e^{(b+1)t}}{(1-e^{t})(b-a+1)} M(t)=(1et)(ba+1)eate(b+1)t
= 1 b − a + 1 u a − u b + 1 u =\frac{1}{b-a+1}\frac{u^a-u^{b+1}}{u} =ba+11uuaub+1
= 1 b − a + 1 1 + a 1 ! ( − u ) + a ( a − 1 ) 2 ! u 2 + a ( a − 1 ) ( a − 2 ) 3 ! ( − u 3 ) + o ( u 3 ) − 1 − b + 1 1 ! ( − u ) − ( b + 1 ) b 2 ! u 2 − ( b + 1 ) b ( b − 1 ) 3 ! ( − u 3 ) − o ( u 3 ) u =\frac{1}{b-a+1}\frac{1+\frac{a}{1!}(-u)+\frac{a(a-1)}{2!}u^2+\frac{a(a-1)(a-2)}{3!}(-u^3)+o(u^3)-1-\frac{b+1}{1!}(-u)-\frac{(b+1)b}{2!}u^2-\frac{(b+1)b(b-1)}{3!}(-u^3)-o(u^3)}{u} =ba+11u1+1!a(u)+2!a(a1)u2+3!a(a1)(a2)(u3)+o(u3)11!b+1(u)2!(b+1)bu23!(b+1)b(b1)(u3)o(u3)
= 1 b − a + 1 a 1 ! ( − u ) + a ( a − 1 ) 2 ! u 2 + a ( a − 1 ) ( a − 2 ) 3 ! ( − u 3 ) + o ( u 3 ) − b + 1 1 ! ( − u ) − ( b + 1 ) b 2 ! u 2 − ( b + 1 ) b ( b − 1 ) 3 ! ( − u 3 ) u =\frac{1}{b-a+1}\frac{\frac{a}{1!}(-u)+\frac{a(a-1)}{2!}u^2+\frac{a(a-1)(a-2)}{3!}(-u^3)+o(u^3)-\frac{b+1}{1!}(-u)-\frac{(b+1)b}{2!}u^2-\frac{(b+1)b(b-1)}{3!}(-u^3)}{u} =ba+11u1!a(u)+2!a(a1)u2+3!a(a1)(a2)(u3)+o(u3)1!b+1(u)2!(b+1)bu23!(b+1)b(b1)(u3)
= 1 b − a + 1 ( ( b + 1 − a ) + a ( a − 1 ) 2 ! u + a ( a − 1 ) ( a − 2 ) 3 ! ( − u 2 ) + o ( u 2 ) − ( b + 1 ) b 2 ! u − ( b + 1 ) b ( b − 1 ) 3 ! ( − u 2 ) ) =\frac{1}{b-a+1}((b+1-a)+\frac{a(a-1)}{2!}u+\frac{a(a-1)(a-2)}{3!}(-u^2)+o(u^2)-\frac{(b+1)b}{2!}u-\frac{(b+1)b(b-1)}{3!}(-u^2)) =ba+11((b+1a)+2!a(a1)u+3!a(a1)(a2)(u2)+o(u2)2!(b+1)bu3!(b+1)b(b1)(u2))
= 1 + a ( a − 1 ) − ( b + 1 ) b 2 ! ( b − a + 1 ) u + ( b + 1 ) b ( b − 1 ) − a ( a − 1 ) ( a − 2 ) 3 ! ( b − a + 1 ) u 2 + o ( u 2 ) =1+\frac{a(a-1)-(b+1)b}{2!(b-a+1)}u+\frac{(b+1)b(b-1)-a(a-1)(a-2)}{3!(b-a+1)}u^2+o(u^2) =1+2!(ba+1)a(a1)(b+1)bu+3!(ba+1)(b+1)b(b1)a(a1)(a2)u2+o(u2)
而 u = 1 − e t = − t − t 2 2 ! + o ( t 2 ) 而u=1-e^t=-t-\frac{t^2}{2!}+o(t^2) u=1et=t2!t2+o(t2)
因 此 M ( t ) = 1 − a ( a − 1 ) − ( b + 1 ) b 2 ! ( b − a + 1 ) t − a ( a − 1 ) − ( b + 1 ) b 2 ! ( b − a + 1 ) t 2 2 ! + ( b + 1 ) b ( b − 1 ) − a ( a − 1 ) ( a − 2 ) 3 ! ( b − a + 1 ) t 2 + o ( t 2 ) 因此M(t)=1-\frac{a(a-1)-(b+1)b}{2!(b-a+1)}t-\frac{a(a-1)-(b+1)b}{2!(b-a+1)}\frac{t^2}{2!}+\frac{(b+1)b(b-1)-a(a-1)(a-2)}{3!(b-a+1)}t^2+o(t^2) M(t)=12!(ba+1)a(a1)(b+1)bt2!(ba+1)a(a1)(b+1)b2!t2+3!(ba+1)(b+1)b(b1)a(a1)(a2)t2+o(t2)
又 因 为 M ( t ) = M ( 0 ) + M ′ ( 0 ) t + M ′ ′ ( 0 ) 2 ! t 2 + o ( t 2 ) 又因为M(t)=M(0)+M'(0)t+\frac{M''(0)}{2!}t^2+o(t^2) M(t)=M(0)+M(0)t+2!M(0)t2+o(t2)
因 此 M ′ ( 0 ) = − a ( a − 1 ) − ( b + 1 ) b 2 ! ( b − a + 1 ) = a + b 2 因此M'(0)=-\frac{a(a-1)-(b+1)b}{2!(b-a+1)}=\frac{a+b}{2} M(0)=2!(ba+1)a(a1)(b+1)b=2a+b
E X = M ′ ( 0 ) = a + b 2 EX=M'(0)=\frac{a+b}{2} EX=M(0)=2a+b
而 M ′ ′ ( 0 ) = 2 ! ∗ ( − a ( a − 1 ) − ( b + 1 ) b 4 ( b − a + 1 ) + ( b + 1 ) b ( b − 1 ) − a ( a − 1 ) ( a − 2 ) 3 ! ( b − a + 1 ) ) 而M''(0)=2!*(-\frac{a(a-1)-(b+1)b}{4(b-a+1)}+\frac{(b+1)b(b-1)-a(a-1)(a-2)}{3!(b-a+1)}) M(0)=2!(4(ba+1)a(a1)(b+1)b+3!(ba+1)(b+1)b(b1)a(a1)(a2))
= a + b 2 + ( b + 1 − a ) ( b 2 + a b − b + a 2 − 2 a ) 3 ( b − a + 1 ) =\frac{a+b}{2}+\frac{(b+1-a)(b^2+ab-b+a^2-2a)}{3(b-a+1)} =2a+b+3(ba+1)(b+1a)(b2+abb+a22a)
= a + b 2 + b 2 + a b − b + a 2 − 2 a 3 =\frac{a+b}{2}+\frac{b^2+ab-b+a^2-2a}{3} =2a+b+3b2+abb+a22a
= 2 a 2 + 2 b 2 + 2 a b + b − a 6 =\frac{2a^2+2b^2+2ab+b-a}{6} =62a2+2b2+2ab+ba
D X = E X 2 − ( E X ) 2 = M ′ ′ ( 0 ) − ( E X ) 2 DX=EX^2-(EX)^2=M''(0)-(EX)^2 DX=EX2(EX)2=M(0)(EX)2
= 2 a 2 + 2 b 2 + 2 a b + b − a 6 − a 2 + 2 a b + b 2 4 =\frac{2a^2+2b^2+2ab+b-a}{6}-\frac{a^2+2ab+b^2}{4} =62a2+2b2+2ab+ba4a2+2ab+b2
= ( b − a + 1 ) 2 − 1 12 =\frac{(b-a+1)^2-1}{12} =12(ba+1)21

2、伯努利分布/两点分布(Bernoulli distribution)

若 X 服 从 伯 努 利 分 布 B ( 1 , p ) , 则 X 满 足 P ( x = 1 ) = p , P ( x = 0 ) = 1 − p = q 若X服从伯努利分布B(1,p) ,则X满足P(x=1)=p, P(x=0)=1-p=q XB(1,p),XP(x=1)=p,P(x=0)=1p=q
M ( t ) = p e t + 1 − p M(t)=pe^{t}+1-p M(t)=pet+1p
φ ( t ) = p e i t + 1 − p \varphi(t)=pe^{it}+1-p φ(t)=peit+1p
M ′ ( t ) = p e t M'(t)=pe^{t} M(t)=pet
E X = M ′ ( 0 ) = p EX=M'(0)=p EX=M(0)=p
M ′ ′ ( t ) = p e t M''(t)=pe^{t} M(t)=pet
E X 2 = M ′ ′ ( 0 ) = p EX^{2}=M''(0)=p EX2=M(0)=p
D X = E X 2 − ( E X ) 2 = p ( 1 − p ) DX=EX^{2}-(EX)^{2}=p(1-p) DX=EX2(EX)2=p(1p)

3、二项分布(Binomial distribution)

若 X 服 从 二 项 分 布 B ( n , p ) , 则 X 满 足 f ( k ; n , p ) = P ( x = k ) = C n k p k ( 1 − p ) n − k ( n 为 整 数 ) 若X服从二项分布B(n,p) ,则X满足f(k;n,p)=P(x=k)=C_{n}^{k}p^k(1-p)^{n-k} (n为整数) XB(n,p),Xf(k;n,p)=P(x=k)=Cnkpk(1p)nk(n)
因 为 服 从 二 项 分 布 的 变 量 可 以 看 作 n 个 独 立 相 同 的 服 从 伯 努 利 分 布 的 变 量 之 和 因为服从二项分布的变量可以看作n个独立相同的服从伯努利分布的变量之和 n
因 此 M ( t ) = ( p e t + 1 − p ) n 因此M(t)=(pe^{t}+1-p)^{n} M(t)=(pet+1p)n
φ ( t ) = ( p e i t + 1 − p ) n \varphi(t)=(pe^{it}+1-p)^{n} φ(t)=(peit+1p)n
M ′ ( t ) = n p ( p e t + 1 − p ) n − 1 e t M'(t)=np(pe^{t}+1-p)^{n-1}e^{t} M(t)=np(pet+1p)n1et
E X = M ′ ( 0 ) = n p EX=M'(0)=np EX=M(0)=np
M ′ ′ ( t ) = n ( n − 1 ) p 2 ( p e t + 1 − p ) n − 2 e 2 t + n p ( p e t + 1 − p ) n − 1 e t M''(t)=n(n-1)p^{2}(pe^{t}+1-p)^{n-2}e^{2t}+np(pe^{t}+1-p)^{n-1}e^{t} M(t)=n(n1)p2(pet+1p)n2e2t+np(pet+1p)n1et
E X 2 = M ′ ′ ( 0 ) = n ( n − 1 ) p 2 + n p EX^{2}=M''(0)=n(n-1)p^{2}+np EX2=M(0)=n(n1)p2+np
D X = E X 2 − ( E X ) 2 = n p ( 1 − p ) DX=EX^{2}-(EX)^{2}=np(1-p) DX=EX2(EX)2=np(1p)

4、几何分布(Geometric distribution)

若 X 服 从 几 何 分 布 G e ( p ) , 则 X 满 足 f ( k ; p ) = P ( x = k ) = ( 1 − p ) k − 1 p ( k = 1 , 2 , 3...... ) 若X服从几何分布Ge(p), 则X满足f(k;p)=P(x=k)=(1-p)^{k-1}p (k=1,2,3......) XGe(p),Xf(k;p)=P(x=k)=(1p)k1p(k=1,2,3......)
M ( t ) = ∑ k = 1 ∞ ( 1 − p ) k − 1 p e t k M(t)=\sum_{k=1}^{\infin}(1-p)^{k-1}pe^{tk} M(t)=k=1(1p)k1petk
= p e t ∑ k = 1 ∞ ( ( 1 − p ) e t ) k − 1 =pe^{t}\sum_{k=1}^{\infin}((1-p)e^t)^{k-1} =petk=1((1p)et)k1
= p e t 1 − ( 1 − p ) e t =\frac{pe^{t}}{1-(1-p)e^{t}} =1(1p)etpet
φ ( t ) = ∑ k = 1 ∞ ( 1 − p ) k − 1 p e i t k \varphi(t)=\sum_{k=1}^{\infin}(1-p)^{k-1}pe^{itk} φ(t)=k=1(1p)k1peitk
= p e i t ∑ k = 1 ∞ ( ( 1 − p ) e i t ) k − 1 =pe^{it}\sum_{k=1}^{\infin}((1-p)e^{it})^{k-1} =peitk=1((1p)eit)k1
= p e i t 1 − ( 1 − p ) e i t =\frac{pe^{it}}{1-(1-p)e^{it}} =1(1p)eitpeit
M ′ ( t ) = p e t ( 1 − ( 1 − p ) e t ) 2 M'(t)=\frac{pe^t}{(1-(1-p)e^t)^2} M(t)=(1(1p)et)2pet
E X = M ′ ( 0 ) = 1 p EX=M'(0)=\frac{1}{p} EX=M(0)=p1
M ′ ′ ( t ) = p e t ( e t − p e t + 1 ) ( 1 − ( 1 − p ) e t ) 3 M''(t)=\frac{pe^t(e^t-pe^t+1)}{(1-(1-p)e^t)^3} M(t)=(1(1p)et)3pet(etpet+1)
E X 2 = M ′ ′ ( 0 ) = 2 − p p 2 EX^{2}=M''(0)=\frac{2-p}{p^2} EX2=M(0)=p22p
D X = E X 2 − ( E X ) 2 = 1 − p p 2 DX=EX^{2}-(EX)^{2}=\frac{1-p}{p^2} DX=EX2(EX)2=p21p

5、负二项分布(Negative binomial distribution)

若 X 服 从 负 二 项 分 布 N B ( r , p ) , 则 X 满 足 f ( k ; r , p ) = ( k + r − 1 k ) p k ( 1 − p ) r , k = 0 , 1 , 2 , 3...... 若X服从负二项分布NB(r,p), 则X满足f(k;r,p)=\binom{k+r-1}{k}p^{k}(1-p)^{r} , k=0,1,2,3...... XNB(r,p),Xf(k;r,p)=(kk+r1)pk(1p)r,k=0,1,2,3......
( r 可 以 为 实 数 , 此 时 的 分 布 称 为 波 利 亚 分 布 ) (r可以为实数,此时的分布称为波利亚分布) (r)
M ( t ) = ∑ k = 0 ∞ ( k + r − 1 k ) p k ( 1 − p ) r e t k M(t)=\sum_{k=0}^{\infin}\binom{k+r-1}{k}p^k(1-p)^re^{tk} M(t)=k=0(kk+r1)pk(1p)retk
= ∑ k = 0 ∞ ( − 1 ) k ( − r k ) p k ( 1 − p ) r e t k =\sum_{k=0}^{\infin}(-1)^k\binom{-r}{k}p^k(1-p)^re^{tk} =k=0(1)k(kr)pk(1p)retk
= ∑ k = 0 ∞ ( − p e t ) k ( − r k ) ( 1 − p ) r =\sum_{k=0}^{\infin}(-pe^t)^k\binom{-r}{k}(1-p)^r =k=0(pet)k(kr)(1p)r
= ( 1 − p ) r ∑ k = 0 ∞ ( − p e t ) k ( − r k ) 1 − r − k =(1-p)^r\sum_{k=0}^{\infin}(-pe^t)^k\binom{-r}{k}1^{-r-k} =(1p)rk=0(pet)k(kr)1rk
= ( 1 − p ) r ( 1 − p e t ) − r =(1-p)^r(1-pe^t)^{-r} =(1p)r(1pet)r
φ ( t ) = ∑ k = 0 ∞ ( k + r − 1 k ) p k ( 1 − p ) r e i t k \varphi(t)=\sum_{k=0}^{\infin}\binom{k+r-1}{k}p^k(1-p)^re^{itk} φ(t)=k=0(kk+r1)pk(1p)reitk
= ∑ k = 0 ∞ ( − 1 ) k ( − r k ) p k ( 1 − p ) r e i t k =\sum_{k=0}^{\infin}(-1)^k\binom{-r}{k}p^k(1-p)^re^{itk} =k=0(1)k(kr)pk(1p)reitk
= ∑ k = 0 ∞ ( − p e i t ) k ( − r k ) ( 1 − p ) r =\sum_{k=0}^{\infin}(-pe^{it})^k\binom{-r}{k}(1-p)^r =k=0(peit)k(kr)(1p)r
= ( 1 − p ) r ∑ k = 0 ∞ ( − p e i t ) k ( − r k ) 1 − r − k =(1-p)^r\sum_{k=0}^{\infin}(-pe^{it})^k\binom{-r}{k}1^{-r-k} =(1p)rk=0(peit)k(kr)1rk
= ( 1 − p ) r ( 1 − p e i t ) − r =(1-p)^r(1-pe^{it})^{-r} =(1p)r(1peit)r
M ′ ( t ) = ( 1 − p ) r ( − r ) ( 1 − p e t ) − r − 1 ( − p e t ) M'(t)=(1-p)^r(-r)(1-pe^{t})^{-r-1}(-pe^t) M(t)=(1p)r(r)(1pet)r1(pet)
= r p ( 1 − p ) r e t ( 1 − p e t ) − r − 1 =rp(1-p)^re^t(1-pe^t)^{-r-1} =rp(1p)ret(1pet)r1
E X = M ′ ( 0 ) = r p 1 − p EX=M'(0)=\frac{rp}{1-p} EX=M(0)=1prp
M ′ ′ ( t ) = r p ( 1 − p ) r e t ( 1 − p e t ) − r − 1 + r p ( 1 − p ) r e t ( − r − 1 ) ( 1 − p e t ) − r − 2 ( − p e t ) M''(t)=rp(1-p)^re^t(1-pe^t)^{-r-1}+rp(1-p)^re^t(-r-1)(1-pe^t)^{-r-2}(-pe^t) M(t)=rp(1p)ret(1pet)r1+rp(1p)ret(r1)(1pet)r2(pet)
E X 2 = r p ( 1 − p ) − 1 + r ( r + 1 ) p 2 ( 1 − p ) − 2 EX^2=rp(1-p)^{-1}+r(r+1)p^2(1-p)^{-2} EX2=rp(1p)1+r(r+1)p2(1p)2
= r p ( 1 − p ) + r ( r + 1 ) p 2 ( 1 − p ) 2 =\frac{rp(1-p)+r(r+1)p^2}{(1-p)^2} =(1p)2rp(1p)+r(r+1)p2
= r p + r 2 p 2 ( 1 − p ) 2 =\frac{rp+r^2p^2}{(1-p)^2} =(1p)2rp+r2p2
D X = E X 2 − ( E X ) 2 = p r ( 1 − p ) 2 DX=EX^2-(EX)^2=\frac{pr}{(1-p)^2} DX=EX2(EX)2=(1p)2pr

6、泊松分布(Poisson distribution)

若 X 服 从 泊 松 分 布 P ( λ ) , 则 P ( X = k ) = e − λ λ k k ! , k = 0 , 1 , 2...... 若X服从泊松分布P(\lambda),则P(X=k)=\frac{e^{-\lambda}\lambda^k}{k!},k=0,1,2...... XP(λ),P(X=k)=k!eλλk,k=0,1,2......
M ( t ) = ∑ k = 0 ∞ e − λ λ k k ! e t k M(t)=\sum_{k=0}^{\infin}\frac{e^{-\lambda}\lambda^k}{k!}e^{tk} M(t)=k=0k!eλλketk
= e − λ ∑ k = 0 ∞ ( λ e t ) k k ! =e^{-\lambda}\sum_{k=0}^{\infin}\frac{(\lambda e^t)^k}{k!} =eλk=0k!(λet)k
= e − λ e λ e t =e^{-\lambda}e^{\lambda e^t} =eλeλet
= e λ ( e t − 1 ) =e^{\lambda (e^t-1)} =eλ(et1)
φ ( t ) = ∑ k = 0 ∞ e − λ λ k k ! e i t k \varphi(t)=\sum_{k=0}^{\infin}\frac{e^{-\lambda}\lambda^k}{k!}e^{itk} φ(t)=k=0k!eλλkeitk
= e − λ ∑ k = 0 ∞ ( λ e i t ) k k ! =e^{-\lambda}\sum_{k=0}^{\infin}\frac{(\lambda e^{it})^k}{k!} =eλk=0k!(λeit)k
= e − λ e λ e i t =e^{-\lambda}e^{\lambda e^{it}} =eλeλeit
= e λ ( e i t − 1 ) =e^{\lambda (e^{it}-1)} =eλ(eit1)
M ′ ( t ) = e λ ( e t − 1 ) λ e t M'(t)=e^{\lambda (e^t-1)}\lambda e^t M(t)=eλ(et1)λet
E X = M ′ ( 0 ) = λ EX=M'(0)=\lambda EX=M(0)=λ
M ′ ′ ( t ) = e λ ( e t − 1 ) λ e t + e λ ( e t − 1 ) λ e t λ e t M''(t)=e^{\lambda (e^t-1)}\lambda e^t+e^{\lambda (e^t-1)}\lambda e^t\lambda e^t M(t)=eλ(et1)λet+eλ(et1)λetλet
E X 2 = M ′ ′ ( 0 ) = λ + λ 2 EX^2=M''(0)=\lambda+\lambda^2 EX2=M(0)=λ+λ2
D X = E X 2 − ( E X ) 2 = λ DX=EX^2-(EX)^2=\lambda DX=EX2(EX)2=λ

三、连续型随机变量的分布

1、连续型均匀分布(Uniform distribution (continuous))

若 X 服 从 连 续 型 均 匀 分 布 U ( a , b ) , 则 f ( x ) = 1 b − a I [ a , b ] ( x ) 若X服从连续型均匀分布U(a,b),则f(x)=\frac{1}{b-a}I_{[a,b]}(x) XU(a,b),f(x)=ba1I[a,b](x)
M ( t ) = ∫ a b 1 b − a e t x d x M(t)=\int_{a}^{b}\frac{1}{b-a}e^{tx}dx M(t)=abba1etxdx
= 1 b − a ∫ a b e t x d x =\frac{1}{b-a}\int_{a}^{b}e^{tx}dx =ba1abetxdx
= 1 b − a ( 1 t e t x ∣ a b ) =\frac{1}{b-a}(\frac{1}{t}e^{tx}\mid_{a}^{b}) =ba1(t1etxab)
= e t b − e t a t ( b − a ) =\frac{e^{tb}-e^{ta}}{t(b-a)} =t(ba)etbeta
φ ( t ) = ∫ a b 1 b − a e i t x d x \varphi(t)=\int_{a}^{b}\frac{1}{b-a}e^{itx}dx φ(t)=abba1eitxdx
= 1 b − a ∫ a b e i t x d x =\frac{1}{b-a}\int_{a}^{b}e^{itx}dx =ba1abeitxdx
= 1 b − a ( 1 i t e i t x ∣ a b ) =\frac{1}{b-a}(\frac{1}{it}e^{itx}\mid_{a}^{b}) =ba1(it1eitxab)
= e i t b − e i t a i t ( b − a ) =\frac{e^{itb}-e^{ita}}{it(b-a)} =it(ba)eitbeita
M ′ ( t ) = 1 b − a ( b e t b − a e t a ) t − ( e t b − e t a ) t 2 M'(t)=\frac{1}{b-a}\frac{(be^{tb}-ae^{ta})t-(e^{tb}-e^{ta})}{t^2} M(t)=ba1t2(betbaeta)t(etbeta)
t = 0 为 M ′ ( t ) 的 可 去 间 断 点 , 补 充 定 义 M ′ ( 0 ) = lim ⁡ t → 0 M ′ ( t ) t=0为M'(t)的可去间断点,补充定义M'(0)=\lim_{t\rightarrow0}M'(t) t=0M(t)M(0)=t0limM(t)
E X = M ′ ( 0 ) = lim ⁡ t → 0 ( b e t b − a e t a ) + ( b 2 e t b − a 2 e t a ) t − ( b e t b − a e t a ) 2 t ( b − a ) EX=M'(0)=\lim_{t\rightarrow0}\frac{(be^{tb}-ae^{ta})+(b^2e^{tb}-a^2e^{ta})t-(be^{tb}-ae^{ta})}{2t(b-a)} EX=M(0)=t0lim2t(ba)(betbaeta)+(b2etba2eta)t(betbaeta)
= lim ⁡ t → 0 ( b 2 e t b − a 2 e t a ) 2 ( b − a ) =\lim_{t\rightarrow0}\frac{(b^2e^{tb}-a^2e^{ta})}{2(b-a)} =t0lim2(ba)(b2etba2eta)
= b 2 − a 2 2 ( b − a ) =\frac{b^2-a^2}{2(b-a)} =2(ba)b2a2
= a + b 2 =\frac{a+b}{2} =2a+b
M ′ ′ ( t ) = 1 b − a ( ( b 2 e t b − a 2 e t a ) t + ( b e t b − a e t a ) − ( b e t b − a e t a ) ) t − 2 ( ( b e t b − a e t a ) t − ( e t b − e t a ) ) t 3 M''(t)=\frac{1}{b-a}\frac{((b^2e^{tb}-a^2e^{ta})t+(be^{tb}-ae^{ta})-(be^{tb}-ae^{ta}))t-2((be^{tb}-ae^{ta})t-(e^{tb}-e^{ta}))}{t^3} M(t)=ba1t3((b2etba2eta)t+(betbaeta)(betbaeta))t2((betbaeta)t(etbeta))
= 1 b − a t 2 ( b 2 e t b − a 2 e t a ) − 2 t ( b e t b − a e t a ) + 2 ( e t b − e t a ) t 3 =\frac{1}{b-a}\frac{t^2(b^2e^{tb}-a^2e^{ta})-2t(be^{tb}-ae^{ta})+2(e^{tb}-e^{ta})}{t^3} =ba1t3t2(b2etba2eta)2t(betbaeta)+2(etbeta)
t = 0 为 M ′ ′ ( t ) 的 可 去 间 断 点 , 补 充 定 义 M ′ ′ ( 0 ) = lim ⁡ t → 0 M ′ ′ ( t ) t=0为M''(t)的可去间断点,补充定义M''(0)=\lim_{t\rightarrow0}M''(t) t=0M(t)M(0)=t0limM(t)
E X 2 = M ′ ′ ( 0 ) = lim ⁡ t → 0 1 b − a t 2 ( b 3 e t b − a 3 e t a ) + 2 t ( b 2 e t b − a 2 e t a ) − 2 t ( b 2 e t b − a 2 e t a ) − 2 ( b e t b − a e t a ) + 2 ( b e t b − a e t a ) 3 t 2 EX^2=M''(0)=\lim_{t\rightarrow0}\frac{1}{b-a}\frac{t^2(b^3e^{tb}-a^3e^{ta})+2t(b^2e^{tb}-a^2e^{ta})-2t(b^2e^{tb}-a^2e^{ta})-2(be^{tb}-ae^{ta})+2(be^{tb}-ae^{ta})}{3t^2} EX2=M(0)=t0limba13t2t2(b3etba3eta)+2t(b2etba2eta)2t(b2etba2eta)2(betbaeta)+2(betbaeta)
= 1 b − a lim ⁡ t → 0 t 2 ( b 3 e t b − a 3 e t a ) 3 t 2 =\frac{1}{b-a}\lim_{t\rightarrow0}\frac{t^2(b^3e^{tb}-a^3e^{ta})}{3t^2} =ba1t0lim3t2t2(b3etba3eta)
= 1 b − a lim ⁡ t → 0 ( b 3 e t b − a 3 e t a ) 3 =\frac{1}{b-a}\lim_{t\rightarrow0}\frac{(b^3e^{tb}-a^3e^{ta})}{3} =ba1t0lim3(b3etba3eta)
= 1 b − a ( b 3 − a 3 ) 3 =\frac{1}{b-a}\frac{(b^3-a^3)}{3} =ba13(b3a3)
= b 2 + a b + a 2 3 =\frac{b^2+ab+a^2}{3} =3b2+ab+a2
D X = E X 2 − ( E X ) 2 = ( b − a ) 2 12 DX=EX^2-(EX)^2=\frac{(b-a)^2}{12} DX=EX2(EX)2=12(ba)2

2、指数分布(Exponential distribution)

若 X 服 从 指 数 分 布 E ( λ ) , 则 f ( x ) = λ e − λ x I [ 0 , + ∞ ) ( x ) 若X服从指数分布E(\lambda),则f(x)=\lambda e^{-\lambda x}I_{[0,+\infin)}(x) XE(λ)f(x)=λeλxI[0,+)(x)
M ( t ) = ∫ 0 + ∞ λ e − λ x e t x d x M(t)=\int_{0}^{+\infin} \lambda e^{-\lambda x}e^{tx}dx M(t)=0+λeλxetxdx
= λ ∫ 0 + ∞ e ( t − λ ) x d x =\lambda \int_{0}^{+\infin} e^{(t-\lambda)x}dx =λ0+e(tλ)xdx
= λ t − λ ( e ( t − λ ) x ∣ 0 + ∞ ) =\frac{\lambda}{t-\lambda}(e^{(t-\lambda)x}\mid_{0}^{+\infin}) =tλλ(e(tλ)x0+)
t < λ 时 , M ( t ) = λ t − λ ( 0 − 1 ) t<\lambda时,M(t)=\frac{\lambda}{t-\lambda}(0-1) t<λM(t)=tλλ(01)
= λ λ − t =\frac{\lambda}{\lambda-t} =λtλ
φ ( t ) = λ λ − i t \varphi(t)=\frac{\lambda}{\lambda-it} φ(t)=λitλ
M ′ ( t ) = λ ( λ − t ) 2 M'(t)=\frac{\lambda}{(\lambda-t)^2} M(t)=(λt)2λ
E X = M ′ ( 0 ) = 1 λ EX=M'(0)=\frac{1}{\lambda} EX=M(0)=λ1
M ′ ′ ( t ) = 2 λ ( λ − t ) 3 M''(t)=\frac{2\lambda}{(\lambda-t)^3} M(t)=(λt)32λ
E X 2 = M ′ ′ ( 0 ) = 2 λ 2 EX^2=M''(0)=\frac{2}{\lambda^2} EX2=M(0)=λ22
D X = E X 2 − ( E X ) 2 = 1 λ 2 DX=EX^2-(EX)^2=\frac{1}{\lambda^2} DX=EX2(EX)2=λ21

3、正态分布(Normal distribution)

若 X 服 从 正 态 分 布 N ( μ , σ 2 ) , 则 f ( x ) = 1 2 π σ e − ( x − μ ) 2 2 σ 2 若X服从正态分布N(\mu,\sigma^2),则f(x)=\frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{(x-\mu)^2}{2\sigma^2}} XN(μ,σ2),f(x)=2π σ1e2σ2(xμ)2
引 理 1 : ∫ − ∞ + ∞ e − t 2 2 d t = 2 π 引理1:\int_{-\infin}^{+\infin}e^{-\frac{t^2}{2}}dt=\sqrt{2\pi} 1+e2t2dt=2π
证 明 : ( ∫ − ∞ + ∞ e − t 2 2 d t ) 2 = ∫ − ∞ + ∞ ∫ − ∞ + ∞ e − x 2 + y 2 2 d x d y 证明:(\int_{-\infin}^{+\infin}e^{-\frac{t^2}{2}}dt)^2=\int_{-\infin}^{+\infin}\int_{-\infin}^{+\infin}e^{-\frac{x^2+y^2}{2}}dxdy (+e2t2dt)2=++e2x2+y2dxdy
= ∫ 0 2 π d θ ∫ 0 + ∞ e − r 2 2 r d r =\int_{0}^{2\pi}d\theta \int_{0}^{+\infin}e^{-\frac{r^2}{2}}rdr =02πdθ0+e2r2rdr
= 2 π ∫ 0 + ∞ e − r 2 2 r d r =2\pi \int_{0}^{+\infin}e^{-\frac{r^2}{2}}rdr =2π0+e2r2rdr
= 2 π ( − e − r 2 2 ∣ 0 + ∞ ) =2\pi (-e^{-\frac{r^2}{2}}\mid_{0}^{+\infin}) =2π(e2r20+)
= 2 π =2\pi =2π
因 此 ∫ − ∞ + ∞ e − t 2 2 d t = 2 π 因此\int_{-\infin}^{+\infin}e^{-\frac{t^2}{2}}dt=\sqrt{2\pi} +e2t2dt=2π
M ( t ) = ∫ − ∞ + ∞ 1 2 π σ e − ( x − μ ) 2 2 σ 2 e t x d x M(t)=\int_{-\infin}^{+\infin}\frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{(x-\mu)^2}{2\sigma^2}}e^{tx}dx M(t)=+2π σ1e2σ2(xμ)2etxdx
= 1 2 π σ ∫ − ∞ + ∞ e − ( x − μ ) 2 2 σ 2 + t x d x =\frac{1}{\sqrt{2\pi}\sigma}\int_{-\infin}^{+\infin}e^{-\frac{(x-\mu)^2}{2\sigma^2}+tx}dx =2π σ1+e2σ2(xμ)2+txdx
令 w = x − μ σ 令w=\frac{x-\mu}{\sigma} w=σxμ
原 式 = 1 2 π ∫ − ∞ + ∞ e − w 2 2 + t ( w σ + μ ) d w 原式=\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{w^2}{2}+t(w\sigma+\mu)}dw =2π 1+e2w2+t(wσ+μ)dw
= e μ t 1 2 π ∫ − ∞ + ∞ e − w 2 2 + t σ w d w =e^{\mu t}\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{w^2}{2}+t\sigma w}dw =eμt2π 1+e2w2+tσwdw
= e μ t 1 2 π ∫ − ∞ + ∞ e − ( w − t σ ) 2 − t 2 σ 2 2 d w =e^{\mu t}\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{(w-t\sigma)^2-t^2\sigma^2}{2}}dw =eμt2π 1+e2(wtσ)2t2σ2dw
= e μ t + t 2 σ 2 2 1 2 π ∫ − ∞ + ∞ e − ( w − t σ ) 2 2 d w =e^{\mu t+\frac{t^2\sigma^2}{2}}\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{(w-t\sigma)^2}{2}}dw =eμt+2t2σ22π 1+e2(wtσ)2dw
= e μ t + t 2 σ 2 2 1 2 π 2 π =e^{\mu t+\frac{t^2\sigma^2}{2}}\frac{1}{\sqrt{2\pi}}\sqrt{2\pi} =eμt+2t2σ22π 12π
= e μ t + t 2 σ 2 2 =e^{\mu t+\frac{t^2\sigma^2}{2}} =eμt+2t2σ2
φ ( t ) = ∫ − ∞ + ∞ 1 2 π σ e − ( x − μ ) 2 2 σ 2 e i t x d x \varphi(t)=\int_{-\infin}^{+\infin}\frac{1}{\sqrt{2\pi}\sigma}e^{-\frac{(x-\mu)^2}{2\sigma^2}}e^{itx}dx φ(t)=+2π σ1e2σ2(xμ)2eitxdx
= 1 2 π σ ∫ − ∞ + ∞ e − ( x − μ ) 2 2 σ 2 + i t x d x =\frac{1}{\sqrt{2\pi}\sigma}\int_{-\infin}^{+\infin}e^{-\frac{(x-\mu)^2}{2\sigma^2}+itx}dx =2π σ1+e2σ2(xμ)2+itxdx
令 w = x − μ σ 令w=\frac{x-\mu}{\sigma} w=σxμ
原 式 = 1 2 π ∫ − ∞ + ∞ e − w 2 2 + i t ( w σ + μ ) d w 原式=\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{w^2}{2}+it(w\sigma+\mu)}dw =2π 1+e2w2+it(wσ+μ)dw
= e i μ t 1 2 π ∫ − ∞ + ∞ e − w 2 2 + i t σ w d w =e^{i\mu t}\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{w^2}{2}+it\sigma w}dw =eiμt2π 1+e2w2+itσwdw
= e i μ t 1 2 π ∫ − ∞ + ∞ e − ( w − i t σ ) 2 + t 2 σ 2 2 d w =e^{i\mu t}\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{(w-it\sigma)^2+t^2\sigma^2}{2}}dw =eiμt2π 1+e2(witσ)2+t2σ2dw
= e i μ t − t 2 σ 2 2 1 2 π ∫ − ∞ + ∞ e − ( w − i t σ ) 2 2 d w =e^{i\mu t-\frac{t^2\sigma^2}{2}}\frac{1}{\sqrt{2\pi}}\int_{-\infin}^{+\infin}e^{-\frac{(w-it\sigma)^2}{2}}dw =eiμt2t2σ22π 1+e2(witσ)2dw
= e i μ t − t 2 σ 2 2 1 2 π 2 π =e^{i\mu t-\frac{t^2\sigma^2}{2}}\frac{1}{\sqrt{2\pi}}\sqrt{2\pi} =eiμt2t2σ22π 12π
= e i μ t − t 2 σ 2 2 =e^{i\mu t-\frac{t^2\sigma^2}{2}} =eiμt2t2σ2
M ′ ( t ) = e μ t + t 2 σ 2 2 ( μ + σ 2 t ) M'(t)=e^{\mu t+\frac{t^2\sigma^2}{2}}(\mu+\sigma^2t) M(t)=eμt+2t2σ2(μ+σ2t)
E X = M ′ ( 0 ) = μ EX=M'(0)=\mu EX=M(0)=μ
M ′ ′ ( t ) = e μ t + t 2 σ 2 2 ( μ + σ 2 t ) 2 + e μ t + t 2 σ 2 2 σ 2 M''(t)=e^{\mu t+\frac{t^2\sigma^2}{2}}(\mu+\sigma^2t)^2+e^{\mu t+\frac{t^2\sigma^2}{2}}\sigma^2 M(t)=eμt+2t2σ2(μ+σ2t)2+eμt+2t2σ2σ2
E X 2 = M ′ ′ ( 0 ) = μ 2 + σ 2 EX^2=M''(0)=\mu^2+\sigma^2 EX2=M(0)=μ2+σ2
D X = E X 2 − ( E X ) 2 = σ 2 DX=EX^2-(EX)^2=\sigma^2 DX=EX2(EX)2=σ2
特 别 地 , X 服 从 标 准 正 态 分 布 N ( 0 , 1 ) 时 特别地,X服从标准正态分布N(0,1)时 ,XN(0,1)
M ( t ) = e t 2 2 M(t)=e^{\frac{t^2}{2}} M(t)=e2t2
φ ( t ) = e − t 2 2 \varphi(t)=e^{-\frac{t^2}{2}} φ(t)=e2t2
E X = 0 , D X = 1 EX=0,DX=1 EX=0,DX=1

4、伽马分布(Gamma distribution)

若 X 服 从 伽 马 分 布 Γ ( α , β ) ( α , β > 0 ) , 则 f ( x ) = β α Γ ( α ) x α − 1 e − β x I ( 0 , + ∞ ) ( x ) 若X服从伽马分布\Gamma(\alpha,\beta)(\alpha,\beta>0),则f(x)=\frac{\beta^\alpha}{\Gamma(\alpha)}x^{\alpha-1}e^{-\beta x}I_{(0,+\infin)}(x) XΓ(α,β)(α,β>0),f(x)=Γ(α)βαxα1eβxI(0,+)(x)
其 中 , Γ ( α ) = ∫ 0 + ∞ t α − 1 e − t d t , α > 0 其中,\Gamma(\alpha)=\int_{0}^{+\infin}t^{\alpha-1}e^{-t}dt,\alpha>0 Γ(α)=0+tα1etdt,α>0
指 数 分 布 E ( λ ) 是 伽 马 分 布 Γ ( 1 , λ ) , χ 2 分 布 χ n 2 是 伽 马 分 布 Γ ( n 2 , 1 2 ) 指数分布E(\lambda)是伽马分布\Gamma(1,\lambda),\chi^2分布\chi^2_n是伽马分布\Gamma(\frac{n}{2},\frac{1}{2}) E(λ)Γ(1,λ),χ2χn2Γ(2n,21)
M ( t ) = ∫ 0 + ∞ β α Γ ( α ) x α − 1 e − β x e t x d x M(t)=\int_{0}^{+\infin}\frac{\beta^\alpha}{\Gamma(\alpha)}x^{\alpha-1}e^{-\beta x}e^{tx}dx M(t)=0+Γ(α)βαxα1eβxetxdx
= ∫ 0 + ∞ β α Γ ( α ) x α − 1 e ( t − β ) x d x =\int_{0}^{+\infin}\frac{\beta^\alpha}{\Gamma(\alpha)}x^{\alpha-1}e^{(t-\beta) x}dx =0+Γ(α)βαxα1e(tβ)xdx
= β α ∫ 0 + ∞ 1 Γ ( α ) x α − 1 e ( t − β ) x d x =\beta^\alpha\int_{0}^{+\infin}\frac{1}{\Gamma(\alpha)}x^{\alpha-1}e^{(t-\beta) x}dx =βα0+Γ(α)1xα1e(tβ)xdx
t < β 时 , 令 v = ( β − t ) x , 原 式 = β α β − t ∫ 0 + ∞ 1 Γ ( α ) ( v β − t ) α − 1 e − v d v t<\beta时,令v=(\beta-t)x,原式=\frac{\beta^\alpha}{\beta-t}\int_{0}^{+\infin}\frac{1}{\Gamma(\alpha)}(\frac{v}{\beta-t})^{\alpha-1}e^{-v}dv t<βv=(βt)x=βtβα0+Γ(α)1(βtv)α1evdv
= ( β β − t ) α 1 Γ ( α ) ∫ 0 + ∞ v α − 1 e − v d v =(\frac{\beta}{\beta-t})^\alpha\frac{1}{\Gamma(\alpha)}\int_{0}^{+\infin}v^{\alpha-1}e^{-v}dv =(βtβ)αΓ(α)10+vα1evdv
= ( β β − t ) α 1 Γ ( α ) Γ ( α ) =(\frac{\beta}{\beta-t})^\alpha\frac{1}{\Gamma(\alpha)}\Gamma(\alpha) =(βtβ)αΓ(α)1Γ(α)
= ( β β − t ) α =(\frac{\beta}{\beta-t})^\alpha =(βtβ)α
φ ( t ) = ( β β − i t ) α \varphi(t)=(\frac{\beta}{\beta-it})^\alpha φ(t)=(βitβ)α
M ′ ( t ) = β α ( β − t ) − α − 1 α M'(t)=\beta^\alpha(\beta-t)^{-\alpha-1}\alpha M(t)=βα(βt)α1α
E X = M ′ ( 0 ) = α β EX=M'(0)=\frac{\alpha}{\beta} EX=M(0)=βα
M ′ ′ ( t ) = β α ( β − t ) − α − 2 α ( α + 1 ) M''(t)=\beta^\alpha(\beta-t)^{-\alpha-2}\alpha(\alpha+1) M(t)=βα(βt)α2α(α+1)
E X 2 = α ( α + 1 ) β 2 EX^2=\frac{\alpha(\alpha+1)}{\beta^2} EX2=β2α(α+1)
D X = E X 2 − ( E X ) 2 = α β 2 DX=EX^2-(EX)^2=\frac{\alpha}{\beta^2} DX=EX2(EX)2=β2α

5、贝塔分布(Beta distribution)

若 X 服 从 贝 塔 分 布 B e ( α , β ) ( α , β > 0 ) , 则 f ( x ) = x α − 1 ( 1 − x ) β − 1 ∫ 0 1 u α − 1 ( 1 − u ) β − 1 d u I ( 0 , 1 ) ( x ) 若X服从贝塔分布\Beta e(\alpha,\beta)(\alpha,\beta>0),则f(x)=\frac{x^{\alpha-1}(1-x)^{\beta-1}}{\int_{0}^{1} u^{\alpha-1}(1-u)^{\beta-1}du}I_{(0,1)}(x) XBe(α,β)(α,β>0)f(x)=01uα1(1u)β1duxα1(1x)β1I(0,1)(x)
= Γ ( α + β ) Γ ( α ) Γ ( β ) x α − 1 ( 1 − x ) β − 1 I ( 0 , 1 ) ( x ) = 1 B ( α , β ) x α − 1 ( 1 − x ) β − 1 I ( 0 , 1 ) ( x ) =\frac{\Gamma(\alpha+\beta)}{\Gamma(\alpha)\Gamma(\beta)}x^{\alpha-1}(1-x)^{\beta-1}I_{(0,1)}(x)=\frac{1}{\Beta(\alpha,\beta)}x^{\alpha-1}(1-x)^{\beta-1}I_{(0,1)}(x) =Γ(α)Γ(β)Γ(α+β)xα1(1x)β1I(0,1)(x)=B(α,β)1xα1(1x)β1I(0,1)(x)
其 中 , B ( α , β ) = ∫ 0 1 u α − 1 ( 1 − u ) β − 1 d u , ( α , β > 0 ) 其中,\Beta(\alpha,\beta)=\int_{0}^{1} u^{\alpha-1}(1-u)^{\beta-1}du,(\alpha,\beta>0) ,B(α,β)=01uα1(1u)β1du,(α,β>0)

M ( t ) = ∫ 0 1 1 B ( α , β ) x α − 1 ( 1 − x ) β − 1 e t x d x M(t)=\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha-1}(1-x)^{\beta-1}e^{tx}dx M(t)=01B(α,β)1xα1(1x)β1etxdx
= ∫ 0 1 1 B ( α , β ) x α − 1 ( 1 − x ) β − 1 ∑ k = 0 ∞ ( t x ) k k ! d x =\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha-1}(1-x)^{\beta-1}\sum_{k=0}^{\infin}\dfrac{(tx)^k}{k!}dx =01B(α,β)1xα1(1x)β1k=0k!(tx)kdx
= ∑ k = 0 ∞ ∫ 0 1 1 B ( α , β ) x α − 1 ( 1 − x ) β − 1 ( t x ) k k ! d x =\sum_{k=0}^{\infin}\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha-1}(1-x)^{\beta-1}\dfrac{(tx)^k}{k!}dx =k=001B(α,β)1xα1(1x)β1k!(tx)kdx
= ∑ k = 0 ∞ ∫ 0 1 1 B ( α , β ) x α + k − 1 ( 1 − x ) β − 1 t k k ! d x =\sum_{k=0}^{\infin}\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha+k-1}(1-x)^{\beta-1}\dfrac{t^k}{k!}dx =k=001B(α,β)1xα+k1(1x)β1k!tkdx
= ∑ k = 0 ∞ ∫ 0 1 x α + k − 1 ( 1 − x ) β − 1 B ( α , β ) t k k ! d x =\sum_{k=0}^{\infin}\int_{0}^{1}\frac{x^{\alpha+k-1}(1-x)^{\beta-1}}{\Beta(\alpha,\beta)}\dfrac{t^k}{k!}dx =k=001B(α,β)xα+k1(1x)β1k!tkdx
= ∑ k = 0 ∞ ∫ 0 1 x α + k − 1 ( 1 − x ) β − 1 d x B ( α , β ) t k k ! =\sum_{k=0}^{\infin}\frac{\int_{0}^{1}x^{\alpha+k-1}(1-x)^{\beta-1}dx} {\Beta(\alpha,\beta)}\dfrac{t^k}{k!} =k=0B(α,β)01xα+k1(1x)β1dxk!tk
= ∑ k = 0 ∞ B ( α + k , β ) B ( α , β ) t k k ! =\sum_{k=0}^{\infin}\frac{\Beta(\alpha +k,\beta)} {\Beta(\alpha,\beta)}\dfrac{t^k}{k!} =k=0B(α,β)B(α+k,β)k!tk
= ∑ k = 0 ∞ Γ ( α + k ) Γ ( β ) Γ ( α + β + k ) Γ ( α ) Γ ( β ) Γ ( α + β ) t k k ! =\sum_{k=0}^{\infin}\frac{\frac{\Gamma(\alpha +k)\Gamma(\beta)}{\Gamma(\alpha+\beta+k)}} {\frac{\Gamma(\alpha)\Gamma(\beta)}{\Gamma(\alpha+\beta)}}\dfrac{t^k}{k!} =k=0Γ(α+β)Γ(α)Γ(β)Γ(α+β+k)Γ(α+k)Γ(β)k!tk
= 1 + ∑ k = 1 ∞ ( ∏ r = 0 k − 1 α + r α + β + r ) t k k ! =1+\sum_{k=1}^{\infin}(\prod_{r=0}^{k-1}\frac{\alpha+r}{\alpha+\beta+r})\frac{t^k}{k!} =1+k=1(r=0k1α+β+rα+r)k!tk
E ( X ) = ∫ 0 1 1 B ( α , β ) x α − 1 ( 1 − x ) β − 1 x d x E(X)=\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha-1}(1-x)^{\beta-1}xdx E(X)=01B(α,β)1xα1(1x)β1xdx
= ∫ 0 1 1 B ( α , β ) x α ( 1 − x ) β − 1 d x =\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha}(1-x)^{\beta-1}dx =01B(α,β)1xα(1x)β1dx
= ∫ 0 1 x α ( 1 − x ) β − 1 d x B ( α , β ) =\frac{\int_{0}^{1}x^{\alpha}(1-x)^{\beta-1}dx}{\Beta(\alpha,\beta)} =B(α,β)01xα(1x)β1dx
= B ( α + 1 , β ) B ( α , β ) =\frac{\Beta(\alpha+1,\beta)}{\Beta(\alpha,\beta)} =B(α,β)B(α+1,β)
= Γ ( α + 1 ) Γ ( β ) Γ ( α + β + 1 ) Γ ( α ) Γ ( β ) Γ ( α + β ) =\frac{\frac{\Gamma(\alpha+1)\Gamma(\beta)}{\Gamma(\alpha+\beta+1)}}{\frac{\Gamma(\alpha)\Gamma(\beta)}{\Gamma(\alpha+\beta)}} =Γ(α+β)Γ(α)Γ(β)Γ(α+β+1)Γ(α+1)Γ(β)
= α α + β =\frac{\alpha}{\alpha+\beta} =α+βα
E ( X 2 ) = ∫ 0 1 1 B ( α , β ) x α − 1 ( 1 − x ) β − 1 x 2 d x E(X^2)=\int_{0}^{1}\frac{1}{\Beta(\alpha,\beta)}x^{\alpha-1}(1-x)^{\beta-1}x^2dx E(X2)=01B(α,β)1xα1(1x)β1x2dx
= ∫ 0 1 x α + 1 ( 1 − x ) β − 1 d x B ( α , β ) =\frac{\int_{0}^{1}x^{\alpha+1}(1-x)^{\beta-1}dx}{\Beta(\alpha,\beta)} =B(α,β)01xα+1(1x)β1dx
= B ( α + 2 , β ) B ( α , β ) =\frac{\Beta(\alpha+2,\beta)}{\Beta(\alpha,\beta)} =B(α,β)B(α+2,β)
= Γ ( α + 2 ) Γ ( β ) Γ ( α + β + 2 ) Γ ( α ) Γ ( β ) Γ ( α + β ) =\frac{\frac{\Gamma(\alpha+2)\Gamma(\beta)}{\Gamma(\alpha+\beta+2)}}{\frac{\Gamma(\alpha)\Gamma(\beta)}{\Gamma(\alpha+\beta)}} =Γ(α+β)Γ(α)Γ(β)Γ(α+β+2)Γ(α+2)Γ(β)
= α ( α + 1 ) ( α + β + 1 ) ( α + β ) =\frac{\alpha(\alpha+1)}{(\alpha+\beta+1)(\alpha+\beta)} =(α+β+1)(α+β)α(α+1)
D ( X ) = E ( X 2 ) − ( E X ) 2 D(X)=E(X^2)-(EX)^2 D(X)=E(X2)(EX)2
= α ( α + 1 ) ( α + β + 1 ) ( α + β ) − α 2 ( α + β ) 2 =\frac{\alpha(\alpha+1)}{(\alpha+\beta+1)(\alpha+\beta)}-\frac{\alpha^2}{(\alpha+\beta)^2} =(α+β+1)(α+β)α(α+1)(α+β)2α2
= α β ( α + β + 1 ) ( α + β ) 2 =\frac{\alpha\beta}{(\alpha+\beta+1)(\alpha+\beta)^2} =(α+β+1)(α+β)2αβ

6、t分布(Student’s t-distribution)

若 X 服 从 自 由 度 为 n 的 t 分 布 t ( x , n ) , 则 f ( x ) = Γ ( n + 1 2 ) n π Γ ( n 2 ) ( 1 + x 2 n ) − n + 1 2 , x ∈ R 若X服从自由度为n的t分布t(x,n),则f(x)=\frac{\Gamma(\frac{n+1}{2})}{\sqrt{n\pi}\Gamma(\frac{n}{2})}(1+\frac{x^2}{n})^{-\frac{n+1}{2}},x∈R Xntt(x,n)f(x)=nπ Γ(2n)Γ(2n+1)(1+nx2)2n+1,xR

7 、F分布(F-distribution)

Logo

旨在为数千万中国开发者提供一个无缝且高效的云端环境,以支持学习、使用和贡献开源项目。

更多推荐