On 9/11/07, Paul Smith <[EMAIL PROTECTED]> wrote:
> The package mratios can perform inferences for ratios of normal means.
> Is there some other package to do the same but with non-normal
> populations. Since I have got large samples, an asymptotic procedure
> would be fine.
Thanks for all replies
Paul,
On 9/11/07, JRG <[EMAIL PROTECTED]> wrote:
> On 11 Sep 2007 at 22:10, Robert A LaBudde wrote:
>
> > I think a ratio of two normals has a Cauchy distribution, which
> > doesn't have a variance (the singularity in the denominator), so the
> > Central Limit theorem does not apply.
> >
>
> The C
On 11 Sep 2007 at 22:10, Robert A LaBudde wrote:
> I think a ratio of two normals has a Cauchy distribution, which
> doesn't have a variance (the singularity in the denominator), so the
> Central Limit theorem does not apply.
>
The Cauchy results if the denominator normal distribution has mean
It even does not have a mean, but this does not
matter: both the numerator and denominator are
asymptotically normal and so the ratio converges (in
distribution) to the ratio of two normals (which
indeed has a Cauchy distribution).
--- Robert A LaBudde <[EMAIL PROTECTED]> wrote:
> I think a ratio
I think a ratio of two normals has a Cauchy distribution, which
doesn't have a variance (the singularity in the denominator), so the
Central Limit theorem does not apply.
I would suggest using bootstrap resampling to make inferences.
At 08:10 PM 9/11/2007, Moshe wrote:
>For large samples you ha
For large samples you have asymptotic normality!
--- Paul Smith <[EMAIL PROTECTED]> wrote:
> Dear All,
>
> The package mratios can perform inferences for
> ratios of normal means.
> Is there some other package to do the same but with
> non-normal
> populations. Since I have got large samples, an
Dear All,
The package mratios can perform inferences for ratios of normal means.
Is there some other package to do the same but with non-normal
populations. Since I have got large samples, an asymptotic procedure
would be fine.
Thanks in advance,
Paul
___
7 matches
Mail list logo