Duncan Murdoch writes:

>On Mon, 13 Oct 2003 10:35:17 +0200, "Caroline Le Gall"
><[EMAIL PROTECTED]> wrote:
>
>>Hello all,
>>
>>Is someone can help me ?
>>If X is distributed as a gaussian distribution with m as mean and s as
>>standard deviation, what is the distribution of 1/X ?
>>Particularly, what is the mean and the standard deviation of 1/X ?
>
>Neither the mean nor the s.d. exist, since the density of X is
>positive at 0.   I don't think the distribution has a standard name.
>You can use standard transformation techniques to work out the density
>or CDF from the Gaussian one.

True enough, though you might be able to get a reasonable approximation
if most of the distribution (say plus or minus two standard deviations)
is bounded well away from zero. You could then use a Taylor series
approximation to the function f(x)=1/x to get

f(x)=f(m)+(x-m)f'(m)+other terms.

We hope that these other terms are insignificant. Then the expectation
is

Ef(x)=E[f(m)+(x-m)f'(m)]=f(m)+(m-m)*f'(m)=1/m

and

Var(f(x))=Var(f(m)+(x-m)f'(m))= (f'(m))^2*Var(x) =s^2/m^4.

My math may be a bit rusty, but someone will jump in if I made a
mistake.

I would expect the distribution to be skewed left, with the skewness
being most evident when the bulk of the distribution creeps closer to
zero.

Steve Simon, [EMAIL PROTECTED], Standard Disclaimer.
The STATS web page has moved to
http://www.childrens-mercy.org/stats.

.
.
=================================================================
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