Thread: Probabilities Dealing with Independent Exponential RVs

1. Probabilities Dealing with Independent Exponential RVs

This has been bothering me for a while now:

Q: If $X_i,\,i=1,2,3$ are independent exponential random variables with rates $\lambda_i,\,i=1,2,3$, find

(a) $\mathbb{P}(X_1

Here's what I figured out so far:

Let $Y=\min(X_2,X_3)$ Then $Y$ is $\mathcal{E}(\lambda_2+\lambda_3)$ and is independent of $X_1$. Now observe that $\mathbb{P}(X_1. We now see that

\begin{aligned}\mathbb{P}(X_1

Now, take note that the distribution of $X_2-x$ and $X_3-x$ given that they are larger than $X_1 = x$ is the same as that of $X_2$ and $X_3$, so

\begin{aligned}\mathbb{P}(X_2

----------------------------------------

At this stage, I'm not sure how to tie it all together to get $\mathbb{P}(X_1. Is it valid for me to state that we should simply get $\mathbb{P}(X_1?

I would appreciate any help with this!

2. I cant comment on your solution but i think there might be an easier way
Spoiler:

you can get the joint pdf easily enough, the probability you want is then

$\displaystyle \intop_{x_1=0}^{x_1=x_2} ~ \intop_{x_2=x_1}^{x_2=x_3} ~ \intop_{x_3=x_2}^{x_3=\infty}f(x1,x2,x3) dx_3 dx_2 dx_1$

I haven't done the integration but i would have thought that it was tractable, given that f() is going to be very simple? If nothing else it could verify your answer

3. Originally Posted by Chris L T521
\begin{aligned}\mathbb{P}(X_2
Are you sure of the second equality ?

Anyway, if you really want to do something along the lines of what you've already done, can't you write it this way :

$P(X_1

But $P(X_1<\min(X_2,X_3)\mid X_2

So finally, $P(X_1

Did it go wrong somewhere ? oO

Springfan25 : there's a problem with your boundaries, since for example $x_3$ appears after having integrated with respect to $x_3$, same for $x_2$.

4. Originally Posted by Moo

But $P(X_1<\min(X_2,X_3)\mid X_2
Are you sure about that one?

it must be true that
$\displaystyle (1)~~~~~~P(X_1< X_2) =P(X_1 X_3)P(X_2 > X_3)$
$\displaystyle (1.1)~~~P(X_1< X_2) =P(X_1 X_3)(1-P(X_2 < X_3))$

it must also be true that

$\displaystyle (2) ~~~P(X_1<\min(X_2,X_3)\mid X_2

You propose
$\displaystyle (3)~~~ P(X_1

(2) and (3) Together imply:
$\displaystyle (4)~~~P(X_1

But i think (4) contradicts (1.1)
$\displaystyle (1.1) ~~~P(X_1< X_2) =P(X_1 X_3)(1-P(X_2 < X_3))$
$\displaystyle (4)~~~~~P(X_1

I can only see 2 ways that (4) could be consistent with (1.1)
either $\displaystyle P(X_2 < X_3) =1$ (not true in general)

or $\displaystyle P(X_1 X_3)$. This is clearly not the case for any given value of x3 (eg, consider x3=0), and therefore (i think?!) not true....but im not sure about that reasoning.

In the unlikely event that i am right about the above, i think the integration approach could still work. I may have the limits wrong but there is a constructable region R (albeit unbounded) on which the integral needs to be done. Here's my second go:

Spoiler:

$\displaystyle \intop_{x_1=0}^{x_1=\infty} \intop_{x_2=x_1}^{x_2=\infty} \intop_{x_3=x_2}^{x_3=\infty}f_{x_1,x_2,x_3} dx_3 dx_2 dx_1
$

Im not sure about those upper limits though. I think it would work for a finite upper limit that was the same for x1, x2, x3; but i dont know if it works with infinity.