topic Re: Phyid and Flag in SAS Programming

Phyid and Flag

SASPhile — Tue, 21 Sep 2010 18:38:37 GMT

The phyid's have flag 'C' or 'P'. How do to display phyid that have flag 'C' and 'P'.

for instance:

Phyid flag
1001 C
1002 P
1003 C
1003 P
1004 C
1004 P

display only 1003 and 1004 as they have flag 'C' and 'P'.

Re: Phyid and Flag

RickM — Tue, 21 Sep 2010 18:43:53 GMT

If flag can only have those two values then you could sort by phyid and flag. Then use by group processing with first.flag and last.flag to exclude observations that are both first and last within a group.

Re: Phyid and Flag

polingjw — Tue, 21 Sep 2010 19:14:37 GMT

[pre]

data test;
input phyid flag $;
datalines;
1001 C
1002 P
1003 C
1003 P
1004 C
1004 P
;
run;

proc sql;
select phyid from test where flag = 'C'
intersect
select phyid from test where flag = 'P';
quit;

[/pre]

Re: Phyid and Flag

SASPhile — Tue, 21 Sep 2010 19:24:00 GMT

Thanks.
Phyid flag
1001 C
1002 P
1003 C
1003 P
1004 C
1004 P

if a phyid has C and P and we need only P to be displayed like 1003 and 1004 only P are displayed.
1001 C
1002 P
1003 P
1004 P

Re: Phyid and Flag

Cynthia_sas — Tue, 21 Sep 2010 19:33:06 GMT

Hi:
Do you want an output dataset or do you want a report??? And if you want a report, what is your destination of interest? If you want a dataset, do you just want to drop the obs for the phyid that has both C and P???
cynthia

Re: Phyid and Flag

SASPhile — Tue, 21 Sep 2010 19:35:53 GMT

Need dataset.If a phyid has a flag 'C' and 'P' i need only that record for 'P'.
1001 C
1002 P
1003 C
1003 P
1004 P
1004 C

the output:
1001 C
1002 P
1003 P
1004 P

Re: Phyid and Flag

RickM — Tue, 21 Sep 2010 19:45:17 GMT

You could do this with two sorts. The first sort by phyid and descending flag. The second sort just by phyid and use the nodupkey option.

Re: Phyid and Flag

chang_y_chung_hotmail_com — Tue, 21 Sep 2010 21:14:37 GMT

@polingjw: Here is another way, taking advantage of proc sql's automatic re-merging of the summary statistics:
[pre]
/* test data by polingjw */
data test;
input phyid flag $;
datalines;
1001 C
1002 P
1003 C
1003 P
1004 C
1004 P
;
run;

proc sql;
create table test2 as
select phyid, flag, sum(flag="C") as c, sum(flag="P") as p
from test
group by phyid;

/* print out an obs if the id has only C or P;
the flag="P" obs if the id has both */
select * from test2
where (c and not p) or (not c and p) or (c and p and flag="P");
/* on lst
phyid flag c p
--------------------------------------
1001 C 1 0
1002 P 0 1
1003 P 1 1
1004 P 1 1
*/
quit;
[/pre]

Re: Phyid and Flag

SASPhile — Wed, 22 Sep 2010 13:17:25 GMT

This is interesting.Thanks for the help.