Steven Taylor wrote:
> Could it be prettier?
> corr=:(+/@:([...@-@(m...@[))@*@(]...@-@(m...@]))) @%:@
> ((+/@:(*:@[...@-@(m@ [)))@*@(+/@:(*:@ (]...@-@(m...@])))))
I like the symmetry betwen +/@:* and *&(+/) in this formulation:
correl =: (+/@:* % *&(+/)&.:*:)&(-m)
But if I were writing a real life application that required correlations, I
would use the standard definition:
load 'stats'
1 2 3 corr_z_ 4 5 6
1
corr_z_ f.
(+/@(*~ (-"_1 _ (+/ % #))) % <:@#...@]) % *&(%:@(+/@:*:@(-"_1 _ (+/ %
#)) %
<:@#))
because I trust Jsoftware has understands statistics better than I do.
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
