[
https://issues.apache.org/jira/browse/MATH-1105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13920851#comment-13920851
]
Gilles commented on MATH-1105:
------------------------------
bq. We now have two least squares impls [...]
Arguably, this has been the case for a long time, as witnessed by the
"AbstractLeastSquaresOptimizer" deprecated implementations (two versions:
"o.a.c.m.optimization.general" and "o.a.c.m.optim.nonlinear.vector.jacobian").
Enhancing a "[least-squares] problem" through wrapping was the ultimate goal of
this major refactoring. This is a very illustrative example of the power of the
approach proposed by Evan.
bq. The stats side of it (residual analysis, ANOVA, etc.) belongs more
naturally in stats.regression.
Do you mean that the decision-making (i.e. how to handle the "bad" data) should
be coded there?
IIUC, then an enhanced "LeastSquareProblem" could still be generated that would
take something like a "BadDataHandler" as an argument.
bq. First patch provides the capability.
Regarding the first patch, is it necessary to introduce a new method
"getAcceptedObservations()"? It would be nicer to consider that the wrapper
transparently modifies the problem (by changing the weights of some
observations, or removing them altogether) so that, viewed from the outside, it
would be like a fresh problem was built in the first place (with the modified
observations).
For example, one would be able to tell that some observations were deleted by
taking the difference between calls to "getObservationSize()" on the original
instance and on the wrapped instance.
Does that make sense?
> Least squares statistical data editing
> --------------------------------------
>
> Key: MATH-1105
> URL: https://issues.apache.org/jira/browse/MATH-1105
> Project: Commons Math
> Issue Type: Improvement
> Reporter: Evan Ward
> Attachments: 0001-Add-statistical-editing-capability.patch,
> 0002-Integrate-data-editing-with-the-LS-framework.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
