Let $X$ and $Y$ be an independent random variables with exponential distribution $Exp(1)$.

Compute $E[\min{\{X,Y\}}|Y^{2}]$.

My attempt:

I. If we want to compute $E[\min{\{X,Y\}}|Y]$ may be sufficient to use the following formula:


Is this correct? Is there any other simpler method to compute this conditional expectation?

II. What about $E[\min{\{X,Y\}}|Y^{2}]$? First thing which comes to my mind is that we should find distribution of vector $(\min{\{X,Y\}},Y^{2})$ and $Y^{2}$ and then use appropriate formula. Am I right? How to find distribution of vector $(\min{\{X,Y\}},Y^{2})$? Is there any simpler method to compute this conditional expecation?

  • 1
    $\begingroup$ HINT: if $X \sim Exp(\lambda_x)$ and $Y \sim Exp(\lambda_y)$, then $Z = \min\{X,Y\} \sim Exp(\lambda_x + \lambda_y)$ $\endgroup$ – Guilherme Thompson Dec 14 '15 at 9:11
  • $\begingroup$ @GuilhermeThompson ?? "HINT" is not useful and "HINT2" is wrong. $\endgroup$ – Did Mar 19 '16 at 10:43
  1. Yes, you are correct. So we have

    $$ \Bbb{E}[\min\{X, Y\} | Y] = \int_{0}^{\infty} \min\{x, Y\} \, e^{-x} \, dx = 1 - e^{-Y}. $$

  2. Notice that both $Y$ and $Y^2$ generate the same $\sigma$-algebra: $\sigma(Y) = \sigma(Y^2)$. So we have

    $$ \Bbb{E}[f(X, Y)|Y^2] = \Bbb{E}[f(X, Y)|\sigma(Y^2)] = \Bbb{E}[f(X, Y)|\sigma(Y)] = \Bbb{E}[f(X, Y)|Y]. $$

    If you have not learned the axiomatic approach to the probability theory, ignore this explanation. We have an alternative, more computational explanation as follows: Let $Z = Y^2$. Then

    \begin{align*} \Bbb{E}[f(X, Y) | Y^2] &= \Bbb{E}[f(X, \sqrt{Z}) | Z] \\ &= \int_{0}^{\infty} f(x, \sqrt{Z}) \, e^{-x} \, dx \\ &= \int_{0}^{\infty} f(x, Y) \, e^{-x} \, dx \\ &= \Bbb{E}[f(X, Y) | Y]. \end{align*}

  • 2
    $\begingroup$ @Leon, Yes, any exponential r.v. is non-negative with probability 1. $\endgroup$ – Sangchul Lee Dec 14 '15 at 9:50
  • 1
    $\begingroup$ I am stuck at a similar problem, would you mind adding a comment on what you did to compute the integral? i.e. $$ \int_0^\infty \min \lbrace x,Y \rbrace e^{-x} dx = ?$$ $\endgroup$ – Spaced Oct 1 '16 at 18:34
  • 1
    $\begingroup$ @Spaced, Have you tried splitting the integral into two parts according to whether $x < Y$ or $x > Y$? $\endgroup$ – Sangchul Lee Oct 1 '16 at 19:58
  • $\begingroup$ Thanks a lot that did the trick! (the best thing being that it's not even a trick, more like an application of the indicator function) $\endgroup$ – Spaced Oct 1 '16 at 20:49

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.