[jira] [Resolved] (CLIMATE-484) Remove rebinning metrics

["Michael Joyce (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CLIMATE-484?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Michael Joyce resolved CLIMATE-484.
-----------------------------------

    Resolution: Fixed

PR Merged!

> Remove rebinning metrics
> ------------------------
>
>                 Key: CLIMATE-484
>                 URL: https://issues.apache.org/jira/browse/CLIMATE-484
>             Project: Apache Open Climate Workbench
>          Issue Type: Task
>          Components: metrics
>    Affects Versions: 0.3-incubating
>            Reporter: Michael Joyce
>            Assignee: Michael Joyce
>              Labels: Newcomer
>             Fix For: 0.4
>
>
> We have a few metrics (Spatial/SeasonalStdDevRatio and similar PatternCorrelation ones) that do a temporal rebin. We've discussed this before, but I think it's worth revisiting.
> In my opinion, we shouldn't be doing rebins/regrids/manipulations during an evaluation unless it's 100% necessary for the metric to make sense. In other words:
> {code}
> ds1 = ocw.local.load_file(...)
> ds2 = ocw.local.load_file(...)
> evaluation = Evaluation.evaluation(ds1, [ds2], [metrics.TemporalStdDevRatio()])
> evaluation.run()
> {code}
> Should really be:
> {code}
> ds1 = ocw.local.load_file(...)
> ds2 = ocw.local.load_file(...)
> ds1 = ocw.dataset_processor.temporal_rebin(ds1, ...)
> ds2 = ocw.dataset_processor.temporal_rebin(ds2, ...)
> evaluation = Evaluation.evaluation(ds1, [ds2], [metrics.StdDevRatio()])
> evaluation.run()
> {code}
> That is a more correct usage of the redesigned API in my opinion.
> Thoughts? Comments? I'm going to send a PR for this change, but let's discuss please.



--
This message was sent by Atlassian JIRA
(v6.2#6252)