Try this one:

DATA test (DROP=c); INPUT A $ B $ Distance; IF (a > b) THEN DO; c = a; a = b; b = c; END; CARDS; Person1 Person2 23 Person2 Person3 40 Person3 Person1 45 Person1 Person4 50 Person4 Person2 60 Person5 Person6 100 Person9 Person5 56 ; RUN; PROC SORT DATA=test; BY a b; RUN;

* --- all possible combinations, but not all distances can be calculated, because no combination of observations exist; PROC SQL; CREATE TABLE temp AS SELECT DISTINCT a FROM test UNION SELECT DISTINCT b FROM test ORDER BY a; QUIT; DATA comb; SET temp NOBS=nobs; IF (_n_ < nobs) THEN DO; DO i=_n_+1 TO nobs; SET temp (RENAME=(a=b)) POINT=i; IF (i < nobs) THEN DO; DO j=i+1 TO nobs; SET temp (RENAME=(a=c)) POINT=j; OUTPUT; END; END; END; END; RUN;

* --- all possible combinations, where distances can be calculated ; DATA comb; LENGTH a b c $8 sumd 8; SET test NOBS=nobs; IF (_n_ < nobs) THEN DO; DO i=_n_+1 TO nobs; SET test (RENAME=(a=a1 b=b1 distance=d1)) POINT=i; IF ((a=a1 AND b^=b1) OR (a^=a1 AND b=b1)) THEN DO; IF (a=a1) THEN c = b1; ELSE c = a1; sumd = distance + d1; OUTPUT; END; END; END; RUN;

On Sat, 7 Mar 1998 15:39:18 -0500, "Charles C. Hinnant" <cchinnan@MAXWELL.SYR.EDU> wrote:

>I have a question regarding generating unique combinations in SAS. > >My data has three variables A , B, and Dist (the distance between the two >variables). > >A B Distance >-- -- --------- >Person1 Person2 23 >Person2 Person3 40 >Person3 Person1 45 >Person1 Person4 50 >Person4 Person2 60 >Person5 person6 100 >Person9 Person5 56 > >Iam trying to find all unique 3 person combinations (groups). > >For instance, in this example the groups would be >(Person1, Person2, Person3) (Person1, Person2, Person4) > >Any ideas would be appreciated...

--------------------------- Andreas Grueninger

PRIVAT: grueni@stuttgart.netsurf.de OFFICIAL: grueninger@lfl.bwl.de ---------------------------