Statistics Review: cmedian (Trac #53)

[scipy-gitbot]

Original ticket http://projects.scipy.org/scipy/ticket/53 on 2006-04-02 by @rkern, assigned to @rkern.

The function cmedian in file source:trunk/Lib/stats/stats.py needs review.

Please look over the StatisticsReview guidelines and add your comments below.

[scipy-gitbot]

@rkern wrote on 2006-04-10

The test for cmedian was changed in ad1b0f2 to address a bugfix in the underlying histogram() routine.

This function, as written, is pretty much obsoleted by the ndarray.median() method. I can't think of a place where it would be preferred.

However, a simple rewriting could make it useful again. Instead of operating on a raw dataset, it could operate on already histogrammed data. I'm sure there are some situations where a histogram is available but the raw data is not. I would suggest, though, that we create a Histogram class that encapsulates all of the information that represents a histogram. Then the histogram versions of cmedian, scoreatpercentile, percentileofscore, and probably a couple others, can be methods on this class.

[josef-pkt]

@rgommers another possible candidate for deprecation
I sent a mail to the dev list

Note: It won't be possible to change the function in a backwards compatible way to follow the last suggestion above, to calculate some statistics directly from histogram data.