Re: Trying to understand Distributions

These data are clearly skewed -- one option is to transform the data with a
power transformation:

log normal is a famous transformation (skewed the other way), but you
can use pretty much anything :-) an exponent > 1 would likely match this
data well.

Here's one random reference I found with a quick Google:

http://seismo.berkeley.edu/~kirchner/eps_120/Toolkits/Toolkit_03.pdf

Coincidentally enough -- that was one of the first hits on Google -- but I
took that class 25 years ago!

-CHB



On Fri, Jul 16, 2021 at 7:06 AM Keith Sloan <[email protected]> wrote:

> Thanks
> On 15/07/2021 18:08, Robert Kern wrote:
>
> On Thu, Jul 15, 2021 at 11:06 AM Keith Sloan <[email protected]>
> wrote:
>
>> Okay I now have three fits
>>
>> I got rid of the clipping on Galaxy Count but gamma and norm still look
>> very similar. Is there another distribution like
>> gamma  but instead of being fatter before the peak is fatter after the
>> peak and would be a better match? i.e in the area
>> between 2.3 and 2.4
>>
> Sure, `johnsonsu` will probably fit. Depending on what you actually want
> to do with the fitted distribution, you might be better served with a
> nonparametric KDE instead.
>
>
> johnsonsu seems to work just fine.
>
> I know there are similar distributions to expon but nor familiar with
>> them. 'Count In Cyl' are integer counts so bin set accordingly
>> but would prefer a distribution where the red line went through the
>> midpoint of each bar of the histogram
>>
> If the data is discrete, you probably want to fit one of the discrete
> distributions instead.
>
>
> https://docs.scipy.org/doc/scipy/reference/stats.html#discrete-distributions
>
> You can start with `geom`, but that first bin looks pretty tall for that.
> Maybe `zipf` or `logser` will work.
>
> Looked at geom and did not see a fit function
>
> Not sure I do want a discrete distribution as what I want to subsequently
> perform is fits for n sets of the three distributions
> and see if I can find any correlation between the distribution fit
> parameters.
>
>
> --
> Robert Kern
>
> _______________________________________________
> SciPy-User mailing [email protected]://mail.python.org/mailman/listinfo/scipy-user
>
> --
> ========== Art & Ceramics ===========https://www.instagram.com/ksloan1952/
>
> _______________________________________________
> SciPy-User mailing list
> [email protected]
> https://mail.python.org/mailman/listinfo/scipy-user
>


-- 

Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

[email protected]

_______________________________________________
SciPy-User mailing list
[email protected]
https://mail.python.org/mailman/listinfo/scipy-user