1.方針
2.7で用いた小技を利用する.今,X>tより,
\begin{align}
x-t&= \int_0^xdy-\int_0^tdy\\
&= \int_t^xdy
\end{align}
である.
1.答案
\begin{align}
r(t)&= \mathbb{E}_{X\sim f_X}[X-t|X\geq t]\\
&= \int_t^\infty(x-t)P[X=x|X\geq t]dx\\
&= \int_t^\infty (x-t)\frac{P[X=x]}{P[X\geq t]}dx\\
&= \int_t^\infty(x-t)\frac{f_X(x)}{1-F_X(t)}dx\\
&= \frac{1}{1-F_X(t)}\int_t^\infty(x-t)f_X(x)dx\\
&= \frac{1}{1-F_X(t)}\int_t^\infty dx\int_t^xdyf_X(x)\\
&= \frac{1}{1-F_X(t)}\int_t^\infty dy\int_y^\infty dxf_X(x)\\
&= \frac{1}{1-F_X(t)}\int_t^\infty\{\int_y^\infty f_X(x)dx\}dy\\
&= \frac{1}{1-F_X(t)}\int_t^\infty 1-F_X(y)dy\\
&= \frac{1}{1-F_X(t)}\int_t^\infty 1-F_X(x)dx
\end{align}
今,指数分布の分布関数は
F_X(x)=1-e^{-\lambda x}
なので,
\begin{align}
r(t)&= \frac{1}{1-(1-e^{-\lambda t})}\int_t^\infty\{1-(1-e^{-\lambda x})\}dx\\
&= \frac{1}{e^{-\lambda t}}\int_t^\infty e^{-\lambda x}dx\\
&= \frac{1}{e^{-\lambda t}}\left[-\frac{1}{\lambda}e^{-\lambda x}\right]_t^\infty\\
&= \frac{1}{\lambda}
\end{align}
2.方針
x^2=2\int_0^xtdt
を用いる.
2.答案
\begin{align}
2\int_0^\infty r(t)\{1-F_X(t)\}dt&= 2\int_0^\infty\frac{1}{1-F_X(t)}\int_t^\infty 1-F_X(x)dx\{1-F_X(t)\}dt\\
&= 2\int_0^\infty dt\int_t^\infty dx\{1-F_X(x)\}\\
&= 2\int_0^\infty dx\int_0^xdt\{1-F_X(x)\}\\
&= 2\int_0^\infty x\{1-F_X(x)\}dx\\
&= \int_0^\infty 2x\{\int_x^\infty f_X(t)dt\}dx\\
&= \int_0^\infty dx\int_x^\infty dt2xf_X(t)\\
&= \int_0^\infty dt\int_0^tdx2xf_X(t)\\
&= \int_0^\infty\{2\int_0^t2xdx\}f_X(t)dt\\
&= \int_0^\infty t^2f_X(t)dt\\
&= \mathbb{E}_{X\sim f_X}[X^2]
\end{align}
参考文献
- 『現代数理統計学の基礎』(久保川達也 著)