0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 3 years have passed since last update.

4.2(易) 確率変数の和の平均,分散

Posted at

1.方針

Wは0か一の値しか取らないので,確率変数:

Z=WX+(1-W)Y

とは,確率pでW,1-pでYの値をとる.

1.答案

Zの分布関数は,

\begin{align}
F_Z(z)&=P[Z\leq z]\\
&= P[WX+(1-W)Y\leq z]\\
&= P[W=1,Z\leq z]+P[W=0,Y\leq z]\\
&= pP[X\leq z]+(1-p)P[Y\leq z]\\
&= pf_X(z)+(1-p)f_Y(z)
\end{align}

3行目=4行目はWがX,Yと独立であることより成立.確率密度関数は,

f_Z(z)=pf_X(z)+(1-p)f_Y(z)

期待値は,

\begin{align}
\mathbb{E}_{Z\sim f_Z}[Z]&= \int zf_Z(z)dz\\
&= \int z\{pf_X(z)+(1-p)f_Y(z)\}dz\\
&= p\int zf_X(z)dz+(1-p)\int zf_Y(z)dz\\
&= p\mu+(1-p)\mu=\mu
\end{align}

分散は,

\begin{align}
Var(Z)&= \int (z-\mu)^2f_Z(z)dz\\
&= \int(z-\mu)^2\{pf_X(z)+(1-p)f_Y(z)\}dz\\
&= p\int(z-\mu)^2f_X(z)dz+(1-p)\int(z-\mu)^2f_Y(z)dz\\
&= p\sigma^2+(1-p)\sigma^2=\sigma^2
\end{align}

2.方針

和の期待値=期待値の和


\mathbb{E}_{X,Y\sim f_{X,Y}}[X+Y]=\mathbb{E}_{X\sim f_X}[X]+\mathbb{E}_{Y\sim f_Y}[Y]

(証明)

\begin{align}
\mathbb{E}_{X,Y\sim f_{X,Y}}[X+Y]&= \int\int(x+y)f_{X,Y}(x,y)dxdy\\
&= \int\int xf_{X,Y}(x,y)dxdy+\int\int yf_{X,Y}(x,y)dxdy\\
&= \int x\{\int f_{X,Y}(x,y)dy\}dx+\int y\{\int f_{X,Y}(x,y)dx\}dy\\
&= \int xf_X(x)dx+\int yf_Y(y)dy
\end{align}

和の分散=分散の和+共分散×2


Var(X+Y)=Var(X)+Var(Y)+2Cov(X,Y)

(証明)

\mu_X=\mathbb{E}_{X\sim f_X}[X],\ \mu_Y=\mathbb{E}_{Y\sim f_Y}[Y]

と置くと,

\begin{align}
Var(X+Y)&= \mathbb{E}_{X,Y\sim f_{X,Y}}[(X+Y)^2]-(\mathbb{E}_{X,Y\sim f_{X,Y}}[X+Y])^2\\
&= \mathbb{E}_{X,Y\sim f_{X,Y}}[X^2+2XY+Y^2]-(\mu_X^2+2\mu_X\mu_Y+\mu_Y^2)\\
&= (\mathbb{E}_{X\sim f_X}[X^2]-\mu_X)^2+(\mathbb{E}_{Y\sim f_Y}[Y^2]-\mu_Y)^2+2\{\mathbb{E}_{X,Y\sim f_{X,Y}}[XY]-\mu_X\mu_Y\}\\
&= Var(X)+Var(Y)+2Cov(X,Y)
\end{align}

問題では,二つ目の分散に関する公式を使う.

答案

Var(U)をwで平方完成して,

\begin{align}
Var(U)&= Var(wX+(1-w)Y)\\
&= \mathbb{E}_{X,Y\sim f_{X,Y}}[\{wX+(1-w)Y\}^2]-\{\mathbb{E}_{X,Y\sim f_{X,Y}}[wX+(1-w)Y]\}^2\\
&= \mathbb{E}_{X,Y\sim f_{X,Y}}[w^2X^2+2w(1-w)XY+(1-w)^2Y^2]-\{w\mu_X+(1-w)\mu_Y\}^2\\
&= w^2(\mathbb{E}_{X\sim f_X}[X^2]-\mu^2)+(1-w)^2(\mathbb{E}_{Y\sim f_Y}[Y^2]-\mu_Y)^2\\
& \ \ \ +2w(1-w)\{\mathbb{E}_{X,Y\sim f_{X,Y}}[XY]-\mu_X\mu_Y\}\\
&= w^2Var(X)+(1-w)^2Var(Y)+2w(1-w)Cov(X,Y)\\
&= w^2\sigma^2+(1-w)^2\sigma^2+2w(1-w)\rho\sigma^2\\
&= 2\sigma-2(1-\rho)(w-\frac{1}{2})^2+\frac{1}{2}(\sigma^2+\rho)
\end{align}

ゆえ,w=1/2の時,Var(U)は最小値:

\frac{1}{2}(\sigma^2+\rho)

を取る.

参考文献

0
0
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?