[math] sum of logs in summary statistics

[lu...@free.fr]

In addition to the statistics required by the StatisticalSummary interface it
implements, the SummaryStatistics class computes the sum of squares and the sum
of logs. It also has setters and getters for the underlying statistics
implementations. However, it does not provide a getSumlg method.

Should the sum of logs computation be deprecated or a getSumlg method added ?

Luc

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org

Re: [math] sum of logs in summary statistics

[Luc Maisonobe <Lu...@free.fr>]

Phil Steitz a écrit :
> On Feb 8, 2008 7:00 AM,  <lu...@free.fr> wrote:
>> luc.maisonobe@free.fr wrote:
>>
>>> In addition to the statistics required by the StatisticalSummary interface it
>>> implements, the SummaryStatistics class computes the sum of squares and the
>>> sum
>>> of logs. It also has setters and getters for the underlying statistics
>>> implementations. However, it does not provide a getSumlg method.
>> The sum of logs is also not used in the equals, hash and toString methods.
>>
>> Luc
>>
>>
>>> Should the sum of logs computation be deprecated or a getSumlg method added ?
>>>
> 
> Interesting.  This is likely a result of refactoring several years
> back when the geometric mean computation used the sum of logs
> instance.  Now it does not, so it is either wasted computation or
> something of value not exposed to the user.  Makes sense to me to add
> getSumLog to SummaryStatistics.  It doesn't need to be included in
> equals or hashcode since geo mean + N equivalence implies log sum
> equivalence.
> 
> Looking again at the code, I now see it as stupid that geometricMean
> in SummaryStatistics does not use the sumOfLogs instance.  If
> geometricMean exposed a setter for its internally wrapped sumOfLogs
> instance, we could just set that in SummaryStatistics and only
> increment the sumOfLogs instance.  It would probably also be an
> improvment for geometricMean to expose a setter for this.
> 
> If there are no objections, I will go ahead and make these changes.
> Thanks for pointing this out, luc.

Go ahead with that.

If you could also have a glimpse at the multivariate summary statistics 
I added yesterday, I would be happy. During my paid work day (in the 
space industry), I am in the process of switching several projects from 
Mantissa to [math] and needed this feature.

I am aware this was really done in a hurry. I have tried to be as 
compliant with univariate statistics as possible, but may have 
completely missed something. I have reused the VectorialCovariance class 
I commited one year ago, but it does not follow the general architecture 
of other statistics. I would really like to have your thoughts about this.

Luc

> 
> Phil
> 
> Phil
> 
> 
>>> Luc
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
>> For additional commands, e-mail: dev-help@commons.apache.org
>>
>>
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
> 
> 



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org

Re: [math] sum of logs in summary statistics

[Phil Steitz <ph...@gmail.com>]

On Feb 8, 2008 7:00 AM,  <lu...@free.fr> wrote:
> luc.maisonobe@free.fr wrote:
>
> > In addition to the statistics required by the StatisticalSummary interface it
> > implements, the SummaryStatistics class computes the sum of squares and the
> > sum
> > of logs. It also has setters and getters for the underlying statistics
> > implementations. However, it does not provide a getSumlg method.
>
> The sum of logs is also not used in the equals, hash and toString methods.
>
> Luc
>
>
> >
> > Should the sum of logs computation be deprecated or a getSumlg method added ?
> >

Interesting.  This is likely a result of refactoring several years
back when the geometric mean computation used the sum of logs
instance.  Now it does not, so it is either wasted computation or
something of value not exposed to the user.  Makes sense to me to add
getSumLog to SummaryStatistics.  It doesn't need to be included in
equals or hashcode since geo mean + N equivalence implies log sum
equivalence.

Looking again at the code, I now see it as stupid that geometricMean
in SummaryStatistics does not use the sumOfLogs instance.  If
geometricMean exposed a setter for its internally wrapped sumOfLogs
instance, we could just set that in SummaryStatistics and only
increment the sumOfLogs instance.  It would probably also be an
improvment for geometricMean to expose a setter for this.

If there are no objections, I will go ahead and make these changes.
Thanks for pointing this out, luc.

Phil

Phil


> > Luc
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org

Re: [math] sum of logs in summary statistics

[lu...@free.fr]

luc.maisonobe@free.fr wrote:

> In addition to the statistics required by the StatisticalSummary interface it
> implements, the SummaryStatistics class computes the sum of squares and the
> sum
> of logs. It also has setters and getters for the underlying statistics
> implementations. However, it does not provide a getSumlg method.

The sum of logs is also not used in the equals, hash and toString methods.

Luc

>
> Should the sum of logs computation be deprecated or a getSumlg method added ?
>
> Luc


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org