1
$\begingroup$

Please note that this question is related to How to find $\mathbb{E}[X\mid\min(X,Y)]$? in which the user Did explains the procedure and ontop of that gives the solution to this question.

Problem: Let $X,Y$ be two independent exponential RV with positive parameters $\lambda, \mu$ respectivly.
Compute $E(X \mid \min(X,Y))$


The solution given from the above link is: $$ E(X \mid \min (X,Y)) =\min(X,Y) + \frac{ \mu}{\lambda+\mu} \frac{1}{\lambda}$$


My approach to the problem: Following a similar routine as in the answer of the link I have already managed to compute $E ( \min(X,Y) \mid X)$ but on this task I fail.

I am aware that $\min(X,Y)$ has exponential distribution with parameter $\lambda+\mu$

If I fix an arbitrary $Z \in L^\infty ( \Omega, \sigma(\min(X,Y)), \mathbb{P})$ I can say that $Z=f(T=\min(X,Y))$ for a bounded borel function $f$ and I want to find $h(t)$ in $E(Xf(T))=E(h(t)f(T))$. Starting: \begin{align}E(Xf(T))&=E(Xf(\min(X,Y)))=E(Xf( \min(X,Y))1_{X<Y})+E(Xf(\min(X,Y))1_{X>Y}) \\ &= E(Xf(X)1_{X<Y})+E(Xf(Y)1_{X>Y}) \\ &\overset{1)}= \int xf(x)1_{x<y}P_X(dx)P_Y(dy) + \int x f(y) 1_{x>y} P_X(dx)P_Y(dy) \\ &\overset{2)}=\int_0^\infty \left( \int_x^\infty P_Y(dy) \right)xf(x)P_X(dx)+ \int_0^\infty \left(\int_y^\infty xP_X(dx) \right)f(y)P_Y(dy) \end{align} Where in 1) I used the fact that $X,Y$ are independent and in 2) Fubini-Lebesgue's Theorem. For the two integrals in braces above I obtain $$\int_x^\infty P_Y(dy) = \int_x^\infty \mu e^{-\mu y}dy= e^{-\mu x} $$ and $$ \int_y^\infty x P_X(dx) = \int_y^\infty x \lambda e^{-\lambda x} dx = \frac{1}{\lambda}e^{-\lambda y} (\lambda y+1) $$ If these calculations are right (I hope they are) then I would obtain that $$E(Xf(T))=E\left(Xf(X)e^{-\mu X} + \frac{1}{\lambda} e^{- \lambda Y}( \lambda Y +1)f(Y) \right) $$

Maybe everything I did was wrong so far, but even if not, I wouldn't know how to take it from here. My next step would be to introduce the indicator function again to obtain a $T= \min(X,Y)$ but then I just get stuck at this line.

Are my calculations wrong? if so, please show me where I went wrong and maybe provide a hint on how I can continue

$\endgroup$
  • $\begingroup$ Allow me to add that I really want to follow the procedure as described in the link provided at the intro of the OP. I am sure that there exists advanced theorems which make my tedious calculations obsolete and reduces this problem to two lines, but I haven't seen any of them yet. $\endgroup$ – Spaced Oct 3 '16 at 19:59
2
$\begingroup$

Continuing your calculation, the right side of your last display can be written as \begin{equation*} \begin{split} \int_0^\infty &\left(tf(t)\lambda e^{-(\mu+\lambda)t}+\mu\lambda^{-1}(\lambda t+1)f(t)e^{-(\mu+\lambda)t}\right)dt=\\ &=\int_0^\infty f(t) [t(\lambda+\mu) +\mu\lambda^{-1}]e^{-(\mu+\lambda)t}dt=\\ &=\frac{1}{\mu+\lambda}\int_0^\infty f(t) [t(\lambda+\mu) +\mu\lambda^{-1}]P_T(dt)=\\ &=\frac{1}{\mu+\lambda}E(f(T)\cdot[T(\lambda+\mu)+\mu\lambda^{-1}]) \end{split} \end{equation*} This means that $E(X|T)=T+{\mu\over\lambda(\mu+\lambda)}$.

$\endgroup$
1
$\begingroup$

Note that the density function of $\min(X, Y)$ is given by $(\mu+\lambda)e^{-(\mu+\lambda)x}$. Moreover, \begin{align*} E\big(Xf(\min(X, Y))\big) &=E\big(\min(X, Y)f(\min(X, Y)) + [X-\min(X, Y)] f(\min(X, Y))\big)\\ &=E\big(\min(X, Y)f(\min(X, Y)) + [X-Y] f(Y)1_{X>Y}\big). \end{align*} Furthermore, \begin{align*} E\big((X-Y) f(Y)1_{X>Y}\big)&=E\Big(E\big((X-Y) f(Y)1_{X>Y}\mid Y\big)\Big)\\ &=E\Big(f(Y)E\big((X-Y) 1_{X>Y}\mid Y\big)\Big)\\ &=E\left(f(Y)\int_Y^{\infty}(x-Y) \lambda e^{-\lambda x}dx \right)\\ &=E\left(f(Y)\frac{1}{\lambda}e^{-\lambda Y}\right)\\ &=\frac{1}{\lambda}\int_0^{\infty}f(y)e^{-\lambda y}\mu e^{-\mu y} dy\\ &=\frac{\mu}{\lambda(\lambda +\mu)}\int_0^{\infty}f(y)(\lambda +\mu)e^{-(\lambda +\mu) y} dy\\ &=\frac{\mu}{\lambda(\lambda +\mu)}E\big(f(\min(X, Y)) \big). \end{align*} That is, \begin{align*} E\big(Xf(\min(X, Y))\big) &=E\left(\min(X, Y)f(\min(X, Y)) + \frac{\mu}{\lambda(\lambda +\mu)}f(\min(X, Y))\right). \end{align*} Therefore, \begin{align*} E\big(X \mid \min(X, Y) \big) = \min(X, Y)+\frac{\mu}{\lambda(\lambda +\mu)}. \end{align*}

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.