[
https://issues.apache.org/jira/browse/MATH-1159?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14180422#comment-14180422
]
Otmar Ertl commented on MATH-1159:
----------------------------------
If multiple quantile values need to be computed simultaneously (e.g lower
quartile, median, and upper quartile), it would make sense to use the P²
algorithm for histograms that uses more than 5 marker positions as described in
the original paper. Depending on the desired quantile values different sets of
marker positions could be used. Using a PSquarePercentile object for each of
the desired quantiles has some overhead. For example, each of them store the
minimum and the maximum value. The 2nd and 4th marker positions used for the
median correspond to the 3rd marker positions of the lower and upper quartiles,
respectively. In summary, a P² algorithm implementation that calculates all
desired quantiles simultaneously would be much more efficient.
> Add quartiles to SummaryStatistics
> ----------------------------------
>
> Key: MATH-1159
> URL: https://issues.apache.org/jira/browse/MATH-1159
> Project: Commons Math
> Issue Type: Improvement
> Affects Versions: 3.3
> Reporter: Phil Steitz
>
> Using PSquarePercentile, we can add quartile computation to
> SummaryStatistics. Since maintaining quartiles will add some overhead,
> implementation should allow the feature to be turned off via some kind of
> constructor flag. This does open the can of worms regarding turning on / off
> other stats, which is probably a good thing to think about as the
> implementation of this feature is developed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
