Xin,
just one more detail. Rafael Perera Salazar, pointed out that:
Rafael Perera-Salazar to me
show details
12:50 pm (2½ hours ago)
Daniel, doesn't that depend on your parametrisation of the gamma
distribution?
What I mean is that in some texts it is defined as
~ Gamma(n, lambda)
while in others this same representation of the density function refers
to
~ Gamma(n,1/lambda)
I think this is why your are getting different results.
That is, be careful with the definition of the Gamma distribution you are
using because the notation is not always standard.
In short, if y ~ Gamma(n, lambda) then ky ~ Gamma(n, lambda/k) provided
that:
f(y,n,lambda) = (lambda/gamma(n)) (lambda*y) ^(n-1) exp(-lambda*y) if y
>= 0
Cheers,
DANIEL
On 3/30/07, Xin <[log in to unmask]> wrote:
>
> Dear All:
>
> Thank you for all of your supporting of my requery!
>
> Finally, I got the final evident using transformation.
>
> If x~Gamma(a,b) and y=kx then
>
> you have y~Gamma(a,b/k) This does not follow the additive property of
> Gamma distribution. It is suprised this is different with Normal
> distribution.
>
> Many Thanks
>
> Xin
> ----- Original Message -----
> From: "Daniel Molinari" <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Thursday, March 29, 2007 10:33 PM
> Subject: Re: Scale of gamma distirbution
>
>
> > Excuse me but for me it is not so evident... If you use the theorem of
> > change of variables, you'll find that if y ~ Gamma(n, lambda), then
> > x= my ~ Gamma(n, lambda/m) (n=degrees of freedom, lambda=frequency of
> > the
> > process)
> >
> > Cheers,
> > DANIEL
> >
> >
> > On 3/28/07, Isaac Dialsingh <[log in to unmask]> wrote:
> >>
> >> Madan Kundu wrote:
> >> > Yes. You are right.
> >> >
> >> > This is quiet evident from the additive property of Gamma
> >> distribution.
> >> >
> >> > Regards
> >> > Madan Gopal Kundu
> >> >
> >> > Xin <[log in to unmask]> wrote:
> >> > Dear All:
> >> >
> >> > If y~Ga(a,b)
> >> >
> >> > What's about the (n-1)y~Ga(a, (n-1)b), is this right?
> >> >
> >> >
> >> > Many Thanks
> >> >
> >> >
> >> >
> >> > --------------
> >> >
> >> >
> >> > Madan Gopal Kundu
> >> > Biostatistician and SAS Programmer
> >> > Ranbaxy R&D
> >> > Gurgaon, Haryana
> >> > India
> >> > Web: http://www.freewebs.com/madanstata
> >> > mobile: 91-9868788406
> >> > e-mail: [log in to unmask]
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > Click to join Statisticians_group
> >> >
> >> > ---------------------------------
> >> > Here's a new way to find what you're looking for - Yahoo! Answers
> >> >
> >> >
> >> Yes. You can use moment generating functions to prove this.
> >>
> >> Isaac
> >>
> >
>
|