commons-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "michael.brzustowicz@gmail.com" <michael.brzustow...@gmail.com>
Subject Re: [math] Summary Stats Higher Moments?
Date Mon, 28 Dec 2015 22:55:59 GMT
Hi Phil,
That would be great! I think adding third, fourth moments and skewness,
kurtosis would be a very useful addition.

Also, considering the formulas in Pebay, perhaps a method like

void merge(SummaryStatistics ss) {
    // use Pebay update formulas to merge un-normalized moments of ss with
"this"
   // if one is singleton use "update" method instead
}

could be added to SummaryStatistics in addition to "update". I realize
AggregateSummaryStatistics takes care of merging 1st,2nd order stats, so
this may be redundant.

If there is something I can do to help, please let me know.
-Mike Brzustowicz

On Wed, Dec 23, 2015 at 5:44 AM, Phil Steitz <phil.steitz@gmail.com> wrote:

> On 12/22/15 9:58 AM, michael.brzustowicz@gmail.com wrote:
> > Hi,
> >
> > I see that
> org.apache.commons.math3.stat.descriptive.DescriptiveStatistics
> > uses the singleton update formulas (from Pebay) for calculating
> > (un-normalized) moments up to the 4th moment. Is there some reason that
> > org.apache.commons.math3.stat.descriptive.SummaryStatistics excludes both
> > third and fourth central moments?
> >
> > Is it just a matter of computational efficiency, ie.
> DescriptiveStatistics
> > calculates moments only when the getter is invoked (and all orders need
> not
> > be calculated at once) while the "storeless" SummaryStatistics would need
> > to calculate all 4 orders at every call to update()?
>
> Yes, that is the reason; but it is really more a matter of no one
> having asked for this feature.  You are correct that the updating
> formulas make this possible and the nested nature of the moments
> means that there should not be much cost to adding the third and
> fourth moments.  I would be happy to review and apply a patch (with
> tests) adding these.
>
> Phil
>
> >  Or is there some other
> > blocker?
> >
> > Thanx,
> > Mike Brzustowicz
> >
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: user-unsubscribe@commons.apache.org
> For additional commands, e-mail: user-help@commons.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message