different criteria for dups

[podarum]

I'm not sure how to title this but here is what I'm looking for..  I have a series of dups but want only the ones that have a USE = Y, BUT if the dups both have USE = N, keep only one (either one).

I tried this but it didn't work :

if ((first.PS = last.PS) and (first.prov NE last.prov)) then first.USE = 'Y';

HAVE:

PS    PROV  USE

H8G    AB      Y

H8G    AB      N

L1T     ON      Y

M1P    ON     Y

R8A    SK      N

R8A    MB     N

for the R8A keep ONLY one of the 2 even if though USE says N.

WANT:

H8G    AB      Y

H8G    AB      N

L1T     ON      Y

M1P    ON     Y

R8A    SK      Y

R8A    MB     N

[Linlin]

proc sort data=have out=want nodupkey;

  by ps use;

run;

[podarum]

Linlin, that singled out the R8A but it did not change it to USE=Y

[Ksharp]

Can you post some more HAVE WANT?

Why the red part will be changed from N to Y?

[podarum]

I would like one of the Dupped N's to cange to Y, because the way they were captured in the earlier steps of data manipulation, where even though R8A is in both Prov's, I want one of the them to be counted, I don't want to lose any PS's.

[Linlin]

proc sort data=have;

by ps use;

run;

data  single dup;

by ps use;

if first.use and last.use then output single;

    else output dup;

proc sort data= dup  nodupkey;

by ps use;

data dup;

    set dup;

use='Y';

run;

data want;

    set single dup;

run;

[podarum]

Great, Thank you