http://www.serpentine.com/blog/2014/06/10/win-bigger-statistical-fights-with-a-better-jackknife/
suggests an interesting technique for finding the largest outlier in a list:
jk=: ((_1 }. 0,+/\) + 1 }. 0,~+/\.) % # - 1:
outlier=: {~ [: (i. >./)@:| jk - (+/%#)
outlier 1 3 2 1
3
(He also suggests a rather complicated technique for minimizing floating
point errors when summing biased numbers. But I can't help but think that
there will be data sets that cause those approaches to break down.)
FYI,
--
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
