Hi Randy,

Sorry, I meant to say: there is NO rule of thumb.

Regards - Jim. >-- >Jim Groeneveld, Netherlands >Statistician/SAS consultant >http://jim.groeneveld.eu.tf

On Fri, 22 Jul 2011 08:48:41 -0400, Jim Groeneveld <jim.1stat@YAHOO.COM> wrote:

>Hi Randy, > >There is rule of thumb; it is dependent of many circumstances. >You just have to test a few different approaches (algorithms) >with a moderate set of data. A (small) data step approach is: > >DATA Flags; > MERGE Events /* assumed sorted by Date, ID_A, ID_B, ID_C */ > Events (RENAME=(Event=Event_Flag) WHERE=(Event_Flag EQ 1)); > BY Date ID_A ID_B ID_C; > IF (MISSING(Event_Flag)) THEN Event_Flag = 0; >RUN; > >This merges Events with a (slightly modified) subset of itself, >while initialising Event_Flag to missing for each new group, >possibly reading its group value for the first record of a group >and retaining Event_Flag values of 1 over a whole group >(group made of unique values for Date, ID_A, ID_B and ID_C). > >This can also be interpreted and explained as merging a single >(or a few) group record(s) multiple times with many group records. > >Regards - Jim. >-- >Jim Groeneveld, Netherlands >Statistician/SAS consultant >http://jim.groeneveld.eu.tf > > >On Fri, 22 Jul 2011 07:21:18 -0400, Randy <randistan69@HOTMAIL.COM> wrote: > >>Dear All: >> The data set that I am working with is very large. Will a data step >>solution be better? >> Randy >> >> >>On Fri, 22 Jul 2011 04:47:33 -0400, Jim Groeneveld <jim.1stat@YAHOO.COM> wrote: >> >>>Hi Randy, >>> >>>There are many solutions possible; one using PROC SQL is: >>> >>>DATA Events; >>> INFILE CARDS TRUNCOVER; >>> INPUT Date $ ID_A $ ID_B $ ID_C $ Event; * (ID_A also read as character); >>> CARDS; >>>01JUL 1 X A >>>01JUL 1 X A >>>01JUL 1 X A >>>01JUL 1 X B >>>01JUL 1 X B 1 >>>01JUL 1 X B >>>01JUL 2 X A >>>01JUL 2 X A >>>01JUL 2 X A >>>02JUL 1 Y C >>>02JUL 1 Y C >>>02JUL 1 Y C 1 >>>; >>>RUN; >>> >>>* Corrections of inconsistencies; >>>* - O1JUL (letter O) changed to 01JUL (digit 0); >>>* - removed Event=1 for 01Jul/2/X/A as in result; >>> >>>TITLE "Dataset = Events"; PROC PRINT DATA=Events; RUN; >>> >>>PROC SQL; >>> CREATE TABLE Flags AS >>> SELECT *, >>> CASE MAX(Event) >>> WHEN . THEN 0 >>> ELSE 1 >>> END AS Flag_Event >>> FROM Events >>> GROUP BY Date, ID_A, ID_B, ID_C; >>>QUIT; >>> >>>TITLE "Dataset = Flags"; PROC PRINT DATA=Flags; RUN; >>> >>>Regards - Jim. >>>-- >>>Jim Groeneveld, Netherlands >>>Statistician, SAS consultant >>>http://jim.groeneveld.eu.tf >>> >>> >>>On Thu, 21 Jul 2011 23:12:25 -0400, Randy <randistan69@HOTMAIL.COM> wrote: >>> >>>>Dear all: >>>> >>>>My data is as follows >>>> >>>>Date ID_A ID_B ID_C Event >>>>O1JUL 1 X A >>>>01JUL 1 X A >>>>01JUL 1 X A >>>>01JUL 1 X B >>>>01JUL 1 X B 1 >>>>01JUL 1 X B >>>>01JUL 2 X A >>>>01JUL 2 X A 1 >>>>01JUL 2 X A >>>>02JUL 1 Y C >>>>02JUL 1 Y C >>>>02JUL 1 Y C 1 >>>> >>>>The data that I want is as follows: >>>> >>>>Date ID_A ID_B ID_C Event Flag_Event >>>>O1JUL 1 X A 0 >>>>01JUL 1 X A 0 >>>>01JUL 1 X A 0 >>>>01JUL 1 X B 1 >>>>01JUL 1 X B 1 1 >>>>01JUL 1 X B 1 >>>>01JUL 2 X A 0 >>>>01JUL 2 X A 0 >>>>01JUL 2 X A 0 >>>>02JUL 1 Y C 1 >>>>02JUL 1 Y C 1 >>>>02JUL 1 Y C 1 1 >>>> >>>>So whenever an event has occurred, I want the event to be flagged for those >>>ID's >>>> >>>> Thanks >>>> Randy