[SciPy-User] scoreatpercentile behaviour
Fri Jan 25 10:05:55 CST 2013
On 25.01.2013 16:17, firstname.lastname@example.org wrote:
> Would this help much if we need the interquartile range, i.e 25 an 75
> (That's the main usecase currently in statsmodels.)
Yes, it would still scale average O(n) with the size of the data. The
interquartile range scales average O(n log n) if we use full sorting.
But keep in mind that O(n) can be slower than O(n log n).
More information about the SciPy-User