topic Re: find record in SAS Procedures

find record

Abraham — Mon, 16 Nov 2015 16:45:00 GMT

Hi,

From the below data, how to get first record based on pid and disease. (should work only on pid with duplicate disease condition)

data abc;
input pid age disease $ country $ sno;
cards;
101 23 sarcoma US 1
101 23 sarcoma US 2
102 43 pneumonia China 1
103 56 syphilis Russia 2
103 52 syphilis Russia 1
103 57 Cpox Russia 3
103 58 Cpox Russia 4
103 59 Cpox Russia 5
103 59 Rpox Russia 6
104 75 Spox Uzbek 1
104 82 Spox Uzbek 2
104 12 Asthma Uzbek 3
104 13 Asthma Uzbek 4
104 14 Asthma Uzbek 5
;

Output

pid	age	disease	country	sno
101	23	sarcoma	US	1
103	52	syphilis	Russia	1
103	57	Cpox	Russia	3
104	75	Spox	Uzbek	1
104	12	Asthma	Uzbek	3

Re: find record

PaigeMiller — Mon, 16 Nov 2015 16:53:50 GMT

UNTESTED CODE

proc sort data=abc;
    by pid disease;
run;
data abc2;
    set abc;
    by pid disease;
    if first.disease;
run;

Re: find record

Astounding — Mon, 16 Nov 2015 17:13:05 GMT

If you are allowed to sort your data, that would be the safest beginning:

proc sort data=have;

by pid disease sno;

run;

After that, you could continue:

data want;

set have;

by pid disease;

if first.disease=1 and last.disease=0;

run;

If you are not allowed to sort your data, you have to assume that they are properly grouped (all records for the same pid/disease appear together, in order). You could then code:

data want;

set have;

by pid disease notsorted;

if first.disease=1 and last.disease=0;

run;

So the result is safer if you are allowed to sort. But it's still obtainable if you can't sort, as long as the data behave.

Good luck.

Re: find record

PaigeMiller — Mon, 16 Nov 2015 17:27:11 GMT

I have never heard of a situation where you are not allowed to sort your data.

Does that actually happen? For what reasons would you not be allowed to sort your data?

Re: find record

Astounding — Mon, 16 Nov 2015 23:52:52 GMT

Paige,

There are situations where you wouldn't want to sort a data set ... size of the data set, existence of indices. But even without good reason, the world of SAS provides many sources of incredible situations. Here are just a few I have either encountered or heard about from others.

One supervisor would not allow a MERGE statement, forcing programmers to use IF/THEN instead. MERGE is just too difficult to master.

Student comments and questions ... well real life is stranger than you could imagine.

x=2;

Student question: Why would you want to do that?

a = b + c;

Student comment/question: You can't add letters.

Sometimes students even produce code that works (or at least generates no errors) but is suitable for a puzzle:

if a = 1 or 2 then b=3 and c=4;

In SAS, as in real life, the truth can be stranger than fiction.

Re: find record

PGStats — Tue, 17 Nov 2015 02:34:03 GMT

To complement @Astounding's ideas, if your records are properly grouped (all records for the same pid/disease appear together) but not always in order, you could do:

data want;
do until(last.disease);
    set abc; by pid disease notsorted;
    firstSno = min(firstSno, sno);
    end;
/* To skip groups with a single record */
if first.disease then call missing(firstSno);
do until(last.disease);
    set abc;  by pid disease notsorted;
    if sno = firstSno then do;
        output;
        /* To get only the first record of ties */
        call missing(firstSno);
        end;
    end;
drop firstSno;
run;