[
https://issues.apache.org/jira/browse/MATH-731?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13184774#comment-13184774
]
Sébastien Brisard commented on MATH-731:
----------------------------------------
Patch committed in {{r1230419}}, with the following changes
* Removed {{getLowerLimit()}} and {{getUpperLimit()}}, which are redundant with
{{getSupportLowerBound()}} and {{getSupportUpperBound()}} (the Javadoc is
explicit enough about that).
* Implemented explicit formulas for {{inverseCumulativeProbability(double)}}.
* Removed DEFAULT_INVERSE_ABSOLUTE_ACCURACY, which has now become meaningless
(since {{inverseCumulativeProbability(double)}} no longer resorts to an
iterative solver). I guess {{getSolverAbsoluteAccuracy()}} should return a
reliable estimate on the absolute accuracy of the explicit calculations
performed in {{inverseCumulativeProbability(double)}}. As I have not a clear
view on what this accuracy is, I've {color:red}arbitrarily{color} initialized
{{solverAbsoluteAccuracy}} with {{FastMath.ulp(c)}}. It should probably rather
be something like {{s * FastMath.ulp(c)}}, with {{s > 1}} depending on {{a}},
{{b}} and {{c}}. {color:red}Thoughts?{color}
* Used default implementation of {{sample()}} instead of proposed version, as
they are *exactly* identical (see explicit implementation of
{{inverseCumulativeProbability(double)}}.
* Removed references to Ivan Frohnes' code.
* Updated serialVersionUID
I'll leave this issue unresolved until we come to an agreement on the accuracy
issue.
Thanks for this contribution!
Afterthoughts: we should probably implement explicit expressions as well for
the inverse cumulative probability of a uniform distribution.
> Triangular distribution
> -----------------------
>
> Key: MATH-731
> URL: https://issues.apache.org/jira/browse/MATH-731
> Project: Commons Math
> Issue Type: New Feature
> Reporter: Dennis Hendriks
> Assignee: Sébastien Brisard
> Labels: distribution
> Fix For: 3.0
>
> Attachments: triangular_distribution.patch
>
>
> It would be nice to have a triangular distribution.
--
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
