1
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 1 year has passed since last update.

4.5 (標準) 二次元確率変数の変数変換1 正規分布の和 (畳み込み,積率母関数)

Last updated at Posted at 2021-10-15

方針

X\sim\mathcal{N}(\mu_1,\sigma_1^2),\ Y\sim\mathcal{N}(\mu_2,\sigma_2^2),\ X,Y:iid

である時,これらの確率変数の和の分布を代表的な2種類の方法で解く.和の分布が有名分布に従う場合は,答案から分かる通り積率母関数から求めるほうが楽である.

答案

まずは畳み込みによって求める.

\begin{cases}
Z=X+Y\\
W=Y
\end{cases}\Leftrightarrow
\begin{cases}
X=Z-W\\
Y=W
\end{cases}

なる一対一変換を考える.

\left|\frac{\partial(X,Y)}{\partial(Z,W)}\right|=\left|\begin{pmatrix}1 & 0\\ -1 & 1\end{pmatrix}\right|=1

より,

\begin{align}
f_{Z,W}(z,w)&= f_{X,Y}(z-w,z)\times 1\\
&= \frac{1}{\sqrt{2\pi\sigma_1^2}}\exp\{-\frac{(z-w-\mu_1)^2}{2\sigma_1^2}\}\frac{1}{\sqrt{2\pi\sigma_2^2}}\exp\{-\frac{(w-\mu_1)^2}{2\sigma_2^2}\}\\
&= \frac{1}{2\pi\sigma_1\sigma_2}\exp\left[-\frac{1}{2}\{\frac{(z-w-\mu_1)^2}{\sigma_1^2}+\frac{(w-\mu_2)^2}{\sigma_2^2}\}\right]\tag{1}
\end{align}

ゆえ,

f_Z(z)=\int f_{Z,W}(z,w)dw\tag{2}

を解けば良い.

(2)式の指数部分の中身をwで平方完成して,

\begin{align}
\frac{(z-w-\mu_1)^2}{\sigma_1^2}+\frac{(w-\mu_2)^2}{\sigma_2^2}\}
&= \frac{w^2+2(\mu_1-z)w+(z-\mu_1)^2}{\sigma_1^2}+\frac{w^2-2\mu_2w+\mu_2^2}{\sigma_2^2}\\
&= (\frac{1}{\sigma_1^2}+\frac{1}{\sigma_2^2})w^2+2(\frac{\mu_1-z}{\sigma_1^2}-\frac{\mu_2}{\sigma_2^2})w+\frac{(z-\mu_1)^2}{\sigma_1^2}+\frac{\mu_2^2}{\sigma_2^2}\\
&= \frac{\sigma_1^2+\sigma_2^2}{\sigma_1^2\sigma_2^2}w^2+2\frac{\sigma_2^2(\mu_1-z)-\sigma_1^2\mu_2}{\sigma_1^2\sigma_2^2}w+\frac{(z-\mu_1)^2}{\sigma_1^2}+\frac{\mu_2^2}{\sigma_2^2}\\
&= \frac{\sigma_1^2+\sigma_2^2}{\sigma_1^2\sigma_2^2}\{w^2+2\frac{\sigma_2^2(\mu_1-z)-\sigma_1^2\mu_2}{\sigma_1^2+\sigma_2^2}w\}+\frac{(z-\mu_1)^2}{\sigma_1^2}+\frac{\mu_2^2}{\sigma_2^2}\\
&= \frac{\sigma_1^2+\sigma_2^2}{\sigma_1^2\sigma_2^2}\left(w+\frac{\sigma_2^2(\mu_1-z)-\sigma_1^2\mu_2}{\sigma_1^2+\sigma_2^2}\right)^2+\frac{(z-\mu_1-\mu_2)^2}{\sigma_1^2+\sigma_2^2}
\end{align}

なので,

\begin{align}
f_{Z,W}(z,w)&\propto \exp\{-\frac{\sigma_1^2+\sigma_2^2}{2\sigma_1^2\sigma_2^2}(w-\frac{\sigma_2^2(z-\mu_1)+\sigma_1^2\mu_2}{\sigma_1^2+\sigma_2^2})\}\\
&\ \ \ \ \times\exp\{-\frac{(z-\mu_1-\mu_2)^2}{2(\sigma_1^2+\sigma_2^2)}\}
\end{align}

以上より,

(1)\propto\exp\{-\frac{(z-\mu_1-\mu_2)^2}{2(\sigma_1^2+\sigma_2^2)}\}

なので,

Z=X+Y\sim\mathcal{N}(\mu_1+\mu_2,\sigma_1^2+\sigma_2^2)

次に,積率母関数によって求める.

\begin{align}
M_{X+Y}(t)&= m_X(t)M_Y(t)\\
&= \exp(\mu_1t+\frac{\sigma_1^2t^2}{2})\exp(\mu_2t+\frac{\sigma_2^2t^2}{2})\\
&= \exp\{(\mu_1+\mu_2)t+\frac{\sigma_1^2+\sigma_2^2)t^2}{2}\}\\
\therefore X+Y&\sim \mathcal{N}(\mu_1+\mu_2,\sigma_1^2+\sigma_2^2)
\end{align}

参考文献

1
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
1
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?