Suppose $X\sim\mathrm{Exp} $ and $Y$ is a non negative independent random variable. Does $P(X-Y>t\mid X>Y)=P(X>t)$, where $t$ is a non negative real number. I think it is true as it seems to be can treat the random variable $Y$ as some constants but not sure whether it is correct.


The claim is correct. This is a generalization of the memoryless property of exponentials.

Observe: $$P(X-Y > t|X>Y) = \int_{0}^{\infty}P(X>y+t|X>y,Y=y)dF_Y(y)$$ $$ = \int_{0}^{\infty}P(X>y+t|X>y)dF_Y(y)$$ due to independence $$ = \int_{0}^{\infty}P(X>t)dF_Y(y)$$ due to memoryless property of exponential $$ = P(X>t)\int_{0}^{\infty}dF_Y(y) = P(X>t)$$


  • $\begingroup$ your first integral should integrate with respect to $y$ instead of $F_Y(y)$, aren't it? $\endgroup$ – Mathematics Dec 21 '12 at 5:24
  • $\begingroup$ If Y has a pdf say $f_y(y)$ then replace $dF_y(y)$ with $f_y(y)dy$. If not, then we have to use the Riemann Stieltjes form which is what I have written. $\endgroup$ – Gautam Shenoy Dec 21 '12 at 12:47

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.