commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark R. Diggory" <mdigg...@latte.harvard.edu>
Subject Re: DO NOT REPLY [Bug 20964] - [math] Remove mode, remove some methods from StatUtils
Date Fri, 20 Jun 2003 21:00:04 GMT
bugzilla@apache.org wrote:

>DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG 
>RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT
><http://nagoya.apache.org/bugzilla/show_bug.cgi?id=20964>.
>ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND 
>INSERTED IN THE BUG DATABASE.
>
>http://nagoya.apache.org/bugzilla/show_bug.cgi?id=20964
>
>[math] Remove mode, remove some methods from StatUtils
>
>
>
>
>
>------- Additional Comments From phil@steitz.com  2003-06-20 20:35 -------
>I want to keep StatUtils as lean as possible, partly because the methods are static.
>  
>
I do agree. Some of the rank stats that got in there were just because I 
was grabbing methods out of AbstractStore and modifying them for the 
static case.

>Regarding the build, all tests run fine for me after the change.  Unfortunately,
>the test cases for UnivariateImpl do not execute the paths that would have
>caused failure. One more reason to get the path coverage up. 
>
Good point, I'll try to come up with a couple more UnivariateImpl cases 
that cover the store case.

>Also, I neglected
>to execute maven clean before test.  My mistake.  I see now that these methods
>are reused, so we have to keep them in StatUtils (unless we decide to drop
>skewness and kurtosis from UnivariateImpl).  I will submit a revised patch
>keeping these things in; but I am getting a little squemish about so many static
>methods. My preference would be to remove skewness and kurtosis from
>UnivariateImpl and StatUtils as well.
>  
>
Ok, I'm always going to be a little touchy about Skew and Kurt, their 
kinda like my first born children. I want to clarify that there are 
perfectly stable UnivariateImpl implementations of skew and kurt in the 
storageless case, they do have a place there, and I have improved them 
to use the same strategy as the "Wests" algorithm for variance that Al 
worked on (they no longer require sum of ^3 and sum of ^4). This is 
thanks to Brents link on recursive calculation of moments (which is 
really what Wests algorithm is).

We could possibly place the storage based implementations into 
UnivariateImpl, but then we're back to replicating the code again 
instead of delegating to a standard implementation.

>I do not want to put the percentile computations into StatUtils. Yes, you are
>correct, the pre-1.2 sort implementation that I embedded in
>abstractStoreUnivariate.getSortedValues should be replaced by the ArrayUtils
>sort.  I will submit a separate patch for this.
>  
>
Ok

-- 

Mark Diggory
Software Developer
Harvard MIT Data Center
http://www.hmdc.harvard.edu



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


Mime
View raw message