> > Take the following probability: P(X1 > a  &  X1+X2 < b)
> >
> > Where X1 and X2 are sums of exponential variables (all with the same
> > rate parameter). So the resulting sums are gamma distributed, right?
> >
> > A first possible way to solve this is by conditioning on one of these
> > variables and doing a lot of the math by hand. However, as the number
> > of variables rises, so does the complexity of the calculations.
> >
> > Is there a way of determining this type of probability numerically?
> > [...]
>
> Since X1 and X2 are positive, the conditions X1 > a and X1+X2 < b
> correspond to the interior of a triangle whose vertices are (a,0),
> (b,0), and (a,b-a). So why not integrate F2(b-x)*f1(x)dx from a to b,
> where f1 is the pdf of X1 and F2 is the cdf of X2. Both f1 and F2
> should be simple function calls, no matter how many variables go into
> X1 and X2. What "math by hand" are you talking about?

So what about:

P(X1>a & X1+X2<b & X2+X3>c)

The example in X1,X2 is a very simple one... The "math by hand" I'm
referring to is conditioning on several variables leading to some lengthy
calculations (splitting up integrals, ...).

thanks,
Miyuko.


.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to