commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Phil Steitz (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (MATH-699) inverseCumulativeDistribution fails with cumulative distribution having a plateau
Date Tue, 08 Nov 2011 03:22:53 GMT

    [ https://issues.apache.org/jira/browse/MATH-699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13146060#comment-13146060
] 

Phil Steitz commented on MATH-699:
----------------------------------

I am sorry, Sebastien.  I am not being very helpful here. As I dig deeper into the code and
archaeology, I am having a hard time seeing how we can really improve the default impl without
knowledge of the underlying distribution.  I agree with you that we should clearly document
its limitations, though, and fix the javadoc contracts everywhere to be correct and consistent.
 The default impl was never really intended to be universal - just a simple solver-based impl
that would work for well-behaved distributions. The normal distribution example above points
to a slight improvement that could now be done.  When that code was written, we had yet to
define the supportLowerBound and supportUpperBound properties.  Now that we have those, the
test in the normal dist case could be moved up into the default impl, using the getters on
the distribution for the return values.

By all means if you can find a way to deal with plateaus or otherwise improve robustness of
the default impl, go for it.  In particular, if you can come up with a better way to set up
the solver, possibly eliminating the domainLower/upperBound methods, that would be great.
 It has always seemed a little ugly to me that we had to implement these methods for every
distribution just so we could get initial guesses for the inverse cum solver.

 
                
> inverseCumulativeDistribution fails with cumulative distribution having a plateau
> ---------------------------------------------------------------------------------
>
>                 Key: MATH-699
>                 URL: https://issues.apache.org/jira/browse/MATH-699
>             Project: Commons Math
>          Issue Type: Bug
>    Affects Versions: 3.0
>            Reporter: S├ębastien Brisard
>            Assignee: S├ębastien Brisard
>            Priority: Minor
>         Attachments: AbstractContinuousDistributionTest.java
>
>
> This bug report follows MATH-692. The attached unit test fails. As required by the definition
in MATH-692, the lower-bound of the interval on which the cdf is constant should be returned.
This is not so at the moment.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

       

Mime
View raw message