分散の基本的性質

分散の基本的性質
\(X\)を確率変数、\(a,b\)を定数とする。

(1)

\[ V\left(X+b\right)=V\left(X\right) \]

(2)

\[ V\left(aX\right)=a^{2}V\left(X\right) \]

(3)

\[ V\left(\sum_{i=1}^{n}a_{i}X_{i}\right)=\sum_{i,j}a_{i}a_{j}Cov\left(X_{i},X_{j}\right) \]

(4)

\[ V\left(a_{1}X_{1}+a_{2}X_{2}\right)=a_{1}{}^{2}V\left(X_{1}\right)+a_{2}{}^{2}V\left(X_{2}\right)+2a_{1}a_{2}Cov\left(X_{1},X_{2}\right) \]

(5)

\[ V(XY)=\left(V\left(X\right)+E^{2}\left(X\right)\right)\left(V\left(Y\right)+E^{2}\left(Y\right)\right)+Cov\left(X^{2},Y^{2}\right)-\left(E\left(X\right)E\left(Y\right)+Cov\left(X,Y\right)\right)^{2} \]
和の分散は標準偏差\(\sigma_{X}\)と相関係数\(\rho_{XY}\)を用いて、
\begin{align*} V\left(\sum_{i=1}^{n}a_{i}X_{i}\right) & =\sum_{i,j}a_{i}a_{j}Cov\left(X_{i},X_{j}\right)\\ & =\sum_{i,j}a_{i}a_{j}\sigma_{X_{i},X_{j}}\\ & =\sum_{i,j}a_{i}a_{j}\sigma_{X_{i}}\sigma_{X_{j}}\frac{\sigma_{X_{i},X_{j}}}{\sigma_{X_{i}}\sigma_{X_{j}}}\\ & =\sum_{i,j}a_{i}a_{j}\sigma_{X_{i}}\sigma_{X_{j}}\rho_{X_{i},X_{j}} \end{align*} とも表される。
また、2つの和は、
\[ V\left(a_{1}X_{1}+a_{2}X_{2}\right)=a_{1}{}^{2}V\left(X_{1}\right))+a_{2}{}^{2}V\left(X_{2}\right)+2a_{1}a_{2}\sigma_{X_{1}}\sigma_{X_{2}}\rho_{X_{1},X_{2}} \] となる。

独立同分布の分散

確率変数\(\left\{ X_{k}\right\} _{k\in\left\{ 1,2,\cdots,n\right\} }\)が全て同じ分布\(\forall k\in\left\{ 1,2,\cdots,n\right\} ,X_{k}=X\)として互いに独立であるとき、
\begin{align*} V\left[\sum_{k=1}^{n}a_{k}X_{k}\right] & =\sum_{i=1}^{n}\sum_{j=1}^{n}a_{i}a_{j}Cov\left[X_{i},X_{j}\right]\\ & =\sum_{i=1}^{n}a_{i}^{2}V\left[X_{i}\right]\\ & =V\left[X\right]\sum_{i=1}^{n}a_{i}^{2} \end{align*} \begin{align*} V\left[\sum_{k=1}^{n}X_{k}\right] & =\sum_{i=1}^{n}\sum_{j=1}^{n}Cov\left[X_{i},X_{j}\right]\\ & =\sum_{i=1}^{n}V\left[X_{i}\right]\\ & =nV\left[X\right] \end{align*} となる。
最初に\(X+X=2X\)とすると、違う結果になるので注意。
互いに独立で同分布な確率変数を独立同分布といいIIDと省略されることもある。

(1)

\begin{align*} V\left(X+b\right) & =E\left(\left(X+b\right)^{2}\right)-E^{2}\left(X+b\right)\\ & =E\left(X^{2}\right)+2bE\left(X\right)+b^{2}-\left(E^{2}\left(X\right)+2bE\left(X\right)+b^{2}\right)\\ & =E\left(X^{2}\right)-E^{2}\left(X\right)\\ & =V\left(X\right) \end{align*}

(2)

\begin{align*} V\left(aX\right) & =E\left(a^{2}X^{2}\right)-E^{2}\left(aX\right)\\ & =a^{2}\left(E\left(X^{2}\right)-E^{2}\left(X\right)\right)\\ & =a^{2}V\left(X\right) \end{align*}

(3)

\begin{align*} V\left(\sum_{i=1}^{n}a_{i}X_{i}\right) & =E\left(\left(\sum_{i=1}^{n}a_{i}X_{i}\right)^{2}\right)-E^{2}\left(\sum_{i=1}^{n}a_{i}X_{i}\right)\\ & =E\left(\sum_{i,j}a_{i}a_{j}X_{i}X_{j}\right)-\left(\sum_{i=1}^{n}a_{i}E\left(X_{i}\right)\right)^{2}\\ & =\sum_{i,j}a_{i}a_{j}E\left(X_{i}X_{j}\right)-\sum_{i,j}a_{i}a_{j}E\left(X_{i}\right)E\left(X_{j}\right)\\ & =\sum_{i,j}a_{i}a_{j}\left(E\left(X_{i}X_{j}\right)-E\left(X_{i}\right)E\left(X_{j}\right)\right)\\ & =\sum_{i,j}a_{i}a_{j}Cov\left(X_{i},X_{j}\right) \end{align*}

(4)

(3)より、
\begin{align*} V\left(a_{1}X_{1}+a_{2}X_{2}\right) & =a_{1}a_{1}Cov\left(X_{1},X_{1}\right)+a_{2}a_{2}Cov\left(X_{2},X_{2}\right)+2a_{1}a_{2}Cov\left(X_{1},X_{2}\right)\\ & =a_{1}{}^{2}V\left(X_{1}\right)+a_{2}{}^{2}V\left(X_{2}\right)+2a_{1}a_{2}Cov\left(X_{1},X_{2}\right) \end{align*}

(5)

\begin{align*} V\left(XY\right) & =E\left(X^{2}Y^{2}\right)-E^{2}\left(XY\right)\\ & =E\left(X^{2}\right)E\left(Y^{2}\right)+Cov\left(X^{2},Y^{2}\right)-\left(E\left(X\right)E\left(Y\right)+Cov\left(X,Y\right)\right)^{2}\\ & =\left(V\left(X\right)+E^{2}\left(X\right)\right)\left(V\left(Y\right))+E^{2}\left(Y\right)\right)+Cov\left(X^{2},Y^{2}\right)-\left(E\left(X\right)E\left(Y\right)+Cov\left(X,Y\right)\right)^{2} \end{align*}

ページ情報
タイトル
分散の基本的性質
URL
https://www.nomuramath.com/vpjoljpj/
SNSボタン