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 3 years have passed since last update.

3.17(易) ガンマ分布

Last updated at Posted at 2021-08-11

方針

絶対値は場合分するよりもまるまる置き換えた方が単純になる場合がある.また指数部分の中に次数高い項があれば大体置換積分する.

ガンマ関数が出てくる分布は置換してガンマ分布に持っていくことが多い.ガンマ分布Ga(alpha,beta)の確率密度関数は

f_X(x)=\frac{\beta^\alpha}{\Gamma(\alpha)}x^{\alpha-1}\exp(-\beta x)\\ \Gamma(\alpha)=(\alpha-1)!=\int_0^\infty t^{\alpha-1}\exp(-t)dt

積率母関数は,

\begin{align}
M_X(t)&=\int_0^\infty\frac{\beta^\alpha}{\Gamma(\alpha)}x^{\alpha-1}\exp\{(t-\beta) x\}dx\\
&= \frac{\beta^\alpha}{\Gamma(\alpha)}{(\beta-t)^\alpha}\int_0^\infty\frac{(\beta-t)^\alpha}{\Gamma(\alpha)}x^{\alpha-1}\exp\{-(\beta-t)x\}dx\\
&= (\frac{\beta}{\beta-t})^\alpha
\end{align}

ガンマ関数を導出する.指数分布Exp(beta)とは,「単位時間あたり平均lambda回発生する事象が,ある時点から次に発生するまでにX単位時間かかる」確率を示すのだった.今,

X_1,...,X_\alpha:iid\sim Exp(\beta)

とすると,その和:

Y=X_1+...+X_\alpha

とは,「単位時間あたり平均lambda回発生する事象が,ある時点から次にalpha回発生するまでにY単位時間かかる」ことを表す確率変数である.

今,指数分布の積率母関数は,

\begin{align}
M_{X_1}(t)&= \mathbb{E}_{X_1\sim Exp(\beta)}[e^{tX_1}]\\
&= \int_0^\infty\beta\exp\{(t-\beta)x\}dx\\
&= \beta\left[\frac{1}{t-\beta}\exp\{(t-\beta)x\}\right]_0^\infty\\
&= \frac{1}{\beta-t}
\end{align}

なので,

\begin{align}
M_Y(t)&= \mathbb{E}_Y[e^{tY}]\\
&= \mathbb{E}_{X_1,...,X_\alpha:iid\sim Exp(\beta)}[e^{tX_1}...e^{tX_\alpha}]\\
&= \mathbb{E}_{X_1\sim Exp(\beta)}[e^{tX_1}]...\mathbb{E}_{X_\alpha\sim Exp(\beta)}[e^{tX_\alpha}]\ \ \ (\because iid)\\
&= (\frac{\beta}{\beta-t})^\alpha
\end{align}

より,これはガンマ分布の積率母関数である.

答案

Y=|X|と置く.Yの分布関数は,

\begin{align}
P[Y\leq y]&= P[|X|\leq y]\\
&= P[-y\leq X\leq y]\\
&= \int_{-\infty}^y f_X(x)dx-\int_{-\infty}^{-y}f_X(x)dx
\end{align}

ゆえ,確率密度関数は

\begin{align}
f_Y(y)&= f_X(y)-f_X(-y)(-1)\\
&= 2C(\alpha)\exp(-y^\alpha),\ \ \ (0<y<\infty)
\end{align}

Yの確率密度関数が全確率1を満たすように正規化定数Cを求める.

\int_0^\infty f_Y(y)dy=\int_0^\infty 2C(\alpha)\exp(-y^\alpha)dy=1

ここで,t=y^alphaと置換すると,

\begin{align}
\int_0^\infty 2C(\alpha)\exp(-y^\alpha)dy&= \int_0^\infty\frac{2C(\alpha)}{\alpha}t^{\frac{1}{\alpha}-1}\exp(-t)dt\\
&= \frac{2C(\alpha)}{\alpha}\Gamma(\frac{1}{\alpha})\int_0^\infty\frac{1}{\Gamma(\frac{1}{\alpha})}t^{\frac{1}{\alpha}-1}\exp(-t)dt\\
&= \frac{2C(\alpha)}{\alpha}\Gamma(\frac{1}{\alpha})=1
\end{align}

より,

C(\alpha)=\frac{\alpha}{2\Gamma(\frac{1}{\alpha})}

次に,

\begin{align}
\mathbb{E}_{X\sim f_X}[|X|^\nu]&= \mathbb{E}_{Y\sim f_Y}[Y^\nu]\\
&= \int_0^\infty\frac{\alpha}{\Gamma(\frac{1}{\alpha})}y^\nu\exp(-y^\alpha)dy
\end{align}

ここで,t=y^alphaと置換すると

\begin{align}
\int_0^\infty\frac{\alpha}{\Gamma(\frac{1}{\alpha})}y^\nu\exp(-y^\alpha)dy&= \int_0^\infty\frac{1}{\Gamma(\frac{1}{\alpha})}t^{\frac{\nu+1}{\alpha}-1}\exp(-t)dt\\
&= \frac{\Gamma(\frac{\nu+1}{\alpha})}{\Gamma(\frac{1}{\alpha})}\int_0^\infty\frac{1}{\Gamma(\frac{\nu+1}{\alpha})}t^{\frac{\nu+1}{\alpha}-1}\exp(-t)dt\\
&= \frac{\Gamma(\frac{\nu+1}{\alpha})}{\Gamma(\frac{1}{\alpha})}
\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?