Thanks mike, your solution is about 2x faster and uses a bit more memory. I need to use ecdfunsorted. Thanks for sharing the additional solution. Each of these solutions will be more than fast enough for what I'm using it for.
On Thu, Aug 14, 2014 at 4:11 AM, [email protected] < [email protected]> wrote: > Fair enough. > I was mainly interested whether you found an improvement > for your needs with the verbs I suggested, especially ecdfsorted. > > > Since you wish to merge the probabilities with your data rows, if your > input data is integer as in the example (unlikely?), it might be worth > keeping the frequencies as integer by not normalising. It's easy > enough to interpret "all rows (with probability) > 0.95" as "all rows > with count > 0.95 * N" > eg > ((#((+/\)))@:(#/.~)) 2 2 2 4 4 6 6 8 > > 3 3 > 3 5 5 7 7 8 > > NB apologies for silly line-wraps in my previous post and > any in this one! > > On 13/08/2014 23:02, Joe Bogner wrote: > > On Wed, Aug > 13, 2014 at 5:56 PM, [email protected] > > <mike_liz.day@tiscali. > co.uk> wrote: > > > >> Isn't it more useful to produce a keyed table of > cumulative > >> frequencies? > >> eg: > >> (~.,.(((%{:)@:(+/\)))@:(#/.~))2 > 2 2 4 4 6 6 8 > > > > Thank you. The keyed table is useful, but ultimately > I need to merge > > into the rows of the source table. I'll be running > queries like -- get > > all rows > 0.95 or < 0.05 in which case I will > want the source table > > with the frequencies as a new column > > > ---------------------------------------------------------------------- > > > For information about J forums see http://www.jsoftware.com/forums.htm > > > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
