Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $X$ be a random variable distributed uniformly over the interval $[0,2]$ and let $Y$ be a random variable distributed exponentially with parameter $1$. Describe a probability space in which $$P(X+Y>2)\geq0.5$$

share|cite|improve this question
up vote 2 down vote accepted

Assume that $X$ is uniformly distributed on $(0,2)$ and define $Y$ as a function of $X$ as follows:

  • If $X\lt1$, then $\mathrm e^{-Y}=\frac12X+\frac12$.
  • If $X\gt1$, then $\mathrm e^{-Y}=\frac12X-\frac12$.

Thus, $\mathrm e^{-Y}$ is uniform on $(\frac12,1)$ in the first case and uniform on $(0,\frac12)$ in the second case. Since $X$ is uniform on $(0,2)$, the two cases are equiprobable, hence $\mathrm e^{-Y}$ is (globally) uniform on $(0,1)$ or, equivalently, $Y$ is exponentially distributed with parameter $1$. Now, it happens that $$ [X+Y\gt2]=[X\gt1]=[X+Y\gt2+\log2], $$ hence there exist some random variables $X$ and $Y$ with the desired distributions such that $\mathbb P(X+Y\geqslant2+\log2)=\frac12$, with $2+\log2=2.69^+$.

To show the claimed property, assume that $X\gt1$, then $X+Y=u(Y)$ with $\mathrm e^{-Y}$ in $(0,\frac12)$ and $u:y\mapsto2\mathrm e^{-y}+1+y$. In particular, $Y\gt\log2$ and the function $u$ is increasing on the interval $(\log2,+\infty)$ hence $u(Y)\gt u(\log2)=2+\log2$ on $[X\gt1]$.

This shows that $[X\gt1]\subseteq[X+Y\gt2+\log2]$, which is enough for our purpose (even though the claimed identity $[X\gt1]=[X+Y\gt2+\log2]$ holds).

Edit: Here is a variant of this construction. Let $a$ in $(0,\frac12)$. Assume once again that $X$ is uniformly distributed on $(0,2)$ and define $Y$ as a function of $X$ as follows:

  • If $X\lt 2a$, then $\mathrm e^{-Y}=\frac12X+1-a$.
  • If $X\gt 2a$, then $\mathrm e^{-Y}=\frac12X-a$.

If $X\gt 2a$, then $X+Y=u_a(Y)$ with $\mathrm e^{-Y}$ in $(0,1-a)$ and $u_a:y\mapsto2\mathrm e^{-y}+2a+y$. In particular, $Y\gt y_a$ with $y_a=-\log(1-a)$ and the function $u_a$ is decreasing on the interval $(y_a,\log2)$ and increasing on the interval $(\log2,+\infty)$, with $u_a(\log2)=1+\log2+2a$. Thus, $u_a(Y)\gt 1+\log2+2a$ on $[X\gt2a]$, and $$ \mathbb P(X+Y\gt 1+\log2+2a)\geqslant 1-a. $$ Thus, every $a$ in $(\frac12(1-\log2),\frac12)$ achieves $\mathbb P(X+Y\geqslant x)\geqslant p$ for some $x\gt2$ and $p\gt0.5$ (our first example being the case $a=\frac12$, for which $x=2+\log2=2.69^+$ and $p=0.5$). For example, $a=\frac12-\frac14\log2$ yields numerically $$ \mathbb P(X+Y\gt\tfrac73)\gt\mathbb P(X+Y\gt2.34)\gt0.67\gt\tfrac23. $$

share|cite|improve this answer
+1 Neat! $\phantom{neat indeed}$ – Sasha Oct 29 '12 at 14:50
This is funny, my box of comments indicates four comments on this answer, but only one seems to have survived... – Did Oct 29 '12 at 19:37
@did I had to retract one too. I was trying to suggest a way for the probability to surpass the value of $1/2$, but then realized by my construction does not have correct marginals. – Sasha Oct 29 '12 at 20:39
@Sasha Your comment made me look at the matter again, see Edit (and thanks). – Did Oct 29 '12 at 21:19
@EvanAad the strategy appears to be that we take $\Omega = [0, 2]$ and $X(\omega) = \omega$; then, for $Y$, we alter slightly the fact that $-\log (1 - \frac X 2) \sim Exp(1)$; starting at $\omega \ge 1$ we glue the part of the exponential near infinty to $\omega = 1$ and the part near $\log(2)$ to $2$; this leads to $Y(\omega) = -\log\frac{\omega - 1}{2}$ for $\omega > 1$ instead of taking $-\log(1 - \frac \omega 2)$ on the whole interval. We are shuffling around the tails of the exponential to move the big part to where we need it ($\omega = 1$) and moving less to $\omega = 2$ where e don't. – guy Oct 29 '12 at 22:35

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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