Take the 2-minute tour ×
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.

Let $\zeta$ and $\eta$ be independent random variable with $\exp(\lambda)$ distribution. What is the distribution of $Z=|\zeta-\eta|$ . I am trying to calculate it by finding $\Pr(\zeta-\eta>x)$, and $\Pr(\eta-\zeta>x)$. Thank you in advance

share|improve this question
1  
Please remove the [probability-theory] tag, as this is an elementary problem. –  cardinal Sep 12 '11 at 18:57
    
@Mike, thanks. Couldn't do it myself. :) –  cardinal Sep 12 '11 at 19:42
    
@cardinal: you're welcome. But surely you could have done it yourself? You only need 500 rep to retag questions. –  Mike Spivey Sep 12 '11 at 20:14
    
@Mike, Strangely enough, I tried and got an error saying I couldn't retag! :) –  cardinal Sep 12 '11 at 21:50
add comment

1 Answer

Since $|\zeta-\eta|\ge0$ almost surely, the probabilities $\Pr(|\zeta-\eta|> x)$ for every nonnegative $x$ fully determine the distribution of $|\zeta-\eta|$. Now, let $x$ be nonnegative. The event $[|\zeta-\eta|>x]$ is the disjoint union of the events $[\zeta-\eta> x]$ and $[\eta-\zeta> x]$. By symmetry, these two events have the same probability, so let us compute the probability of the first one.

For every $y$, $\Pr(\zeta>y)=\mathrm e^{-\lambda y}$, hence, by independence of $\eta$ and $\zeta$, $$ \Pr(\zeta-\eta>x\mid\eta)=\mathrm e^{-\lambda (x+\eta)}. $$ Integrating this with respect to the distribution of $\eta$ yields $$ \Pr(\zeta-\eta>x)=\mathrm e^{-\lambda x}\,\mathrm E(\mathrm e^{-\lambda \eta})=\mathrm e^{-\lambda x}\int\limits_0^{+\infty}\mathrm e^{-\lambda t}\lambda \mathrm e^{-\lambda t}\text{d}t=\frac12\mathrm e^{-\lambda x}. $$ Summing up the contributions of $[\zeta-\eta> x]$ and $[\eta-\zeta> x]$, one gets $\Pr(|\zeta-\eta|>x)=\mathrm e^{-\lambda x}$ for every nonnegative $x$, hence $|\zeta-\eta|$ is exponential with parameter $\lambda$.

Edit Combining the second and third items here, one gets the result above.

share|improve this answer
    
so with the absolute sign, this difference will not be a laplacian distribution? –  bear Sep 11 '11 at 22:14
    
If X has a centered Laplace distribution, the distribution of |X| is exponential. en.wikipedia.org/wiki/… –  Did Sep 11 '11 at 22:21
    
Do we need to include the case for which x<0, $Pr(\zeta-\eta>x)=1-P(\zeta-\eta>-x)=1-\frac{e^(-\lambda x)}{2}$ –  bear Sep 11 '11 at 22:39
    
No, we do not. See edited first paragraph. –  Did Sep 11 '11 at 22:54
add comment

Your Answer

 
discard

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.