[SciPy-User] Gamma distribution in scipy.stats

[josef.pktd at gmail.com]

On Thu, Jan 28, 2010 at 3:22 PM, Gökhan Sever <gokhansever at gmail.com> wrote:
>
>
> On Thu, Jan 28, 2010 at 1:07 PM, Bruce Southey <bsouthey at gmail.com> wrote:
>>
>> On 01/28/2010 12:16 PM, josef.pktd at gmail.com wrote:
>>
>> On Thu, Jan 28, 2010 at 1:08 PM, Gökhan Sever <gokhansever at gmail.com>
>> wrote:
>>
>>
>> On Thu, Jan 28, 2010 at 11:50 AM, Robert Kern <robert.kern at gmail.com>
>> wrote:
>>
>>
>> On Thu, Jan 28, 2010 at 11:36, Gökhan Sever <gokhansever at gmail.com> wrote:
>>
>>
>> Hello,
>>
>> Could someone explain to me why doesn't scipy explicitly use the
>> location
>> and scaling parameters representing its PDF?
>>
>>
>> Because the transformation for the location and scale parameters are
>> the same for every PDF and is well known. However, including them in
>> the formula often clutters it up and obscures the differences between
>> PDFs.
>>
>>
>>
>> GSL and R doesn't use the location parameter then. And Numpy's Gamma PDF
>> includes scaling in the formulae
>>
>> (http://docs.scipy.org/doc/numpy/reference/generated/numpy.random.gamma.html#numpy.random.gamma).
>> I am guessing that everyone has its own style when it comes to represent
>> the
>> distributions. Not so surprisingly it is shown in a different form in my
>> Cloud and Precipitation Parametrization book.
>>
>> I suggest to add a statement like: Gamma distribution is mainly used to
>> represent precipitation distribution in bulk cloud parametrization
>> schemes.
>>
>>
>> I don't think "mainly" is a correct description,
>>
>> a random quote after a short google search
>> "The Gamma distribution is widely used in engineering, science, and
>> business, to model continuous variables that are always positive and
>> have skewed distributions"
>>
>> It's a pretty common distribution.
>>
>> Josef
>>
>>
>>
>>
>> I think what  Gökhan is getting at is that limited description provided in
>> the documentation link:
>> "The Gamma distribution is often used to model the times to failure of
>> electronic components, and arises naturally in processes for which the
>> waiting times between Poisson distributed events are relevant."
>>
>> I actually think that part should be removed from the documentation.
>>
>> Bruce
>>
>
> It doesn't really too much matter to me to include some extra information or
> not to this description, but for the sake of consistency it might be a
> better idea to leave this description from the Gamma distribution page.
> Since neither it nor my proposed addition properly generalizes the use of
> the distribution. Additionally, if such examples exist in other
> distributions they should be removed as well.

this reminded me of a thread a while ago where I looked this up:

For a given Poisson arrival process, the time between two arrivals is
exponentially distributed, the time between k arrivals is gamma
distributed.

For many distribution, there are tables how different distributions
are linked. I don't know whether some of this would be useful
information in the docs. In many cases a quick look on Wikipedia is
very informative about common application and the relationship between
distributions.

Josef


>>
>>
>>
>> _______________________________________________
>> SciPy-User mailing list
>> SciPy-User at scipy.org
>> http://mail.scipy.org/mailman/listinfo/scipy-user
>>
>
>
>
> --
> Gökhan
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>