Selecting observations based on closest records of days

[chuck33]

Hi, I have data as follows

data have;

input subject studyday visit trt_changeday ;

datalines;

1 23 1 30

1 36 2 30

1 40 3 30

1 55 4 30

1 78 5 30

2 23 1 35

2 25 2 35

2 44 3 35

3 12 1 24

3 18 2 24

3 30 3 24

3 45 4 24

3 66 5 24

4 2 1 47

4 56 2 47

4 89 3 47

;

run;

data have;

set have;

gap=trt_changeday-studyday;

run;

I want to flag records based on following conditions

1. if studyday is +30 or -30 days  of trt_changeday then flag the record closest to trt_changeday
else ;

2. if studyday is out of range of +/ -30 days of trt_changeday then flag the record of the visit erlier to trt_changeday (Subject 4)

I need final output as below.

subject	studyday	visit	trt changeday	gap	flag
1	23	1	30	7
1	36	2	30	-6	1
1	40	3	30	-10
1	55	4	30	-25
1	78	5	30	-48
2	23	1	35	12
2	25	2	35	10
2	44	3	35	-9	1
3	12	1	24	12
3	18	2	24	6
3	30	3	24	-6	1
3	45	4	24	-21
3	66	5	24	-42
4	2	1	47	45	1
4	56	2	47	-9
4	89	3	47	-42

Thanks for the help.

[Ksharp · Accepted solution]

It would be better if you could post some more data.

data have;
input subject studyday visit trt_changeday ;
datalines;
1 23 1 30
1 36 2 30
1 40 3 30
1 55 4 30
1 78 5 30
2 23 1 35
2 25 2 35
2 44 3 35
3 12 1 24
3 18 2 24
3 30 3 24
3 45 4 24
3 66 5 24
4 2 1 47
4 56 2 47
4 89 3 47
;
run;
 
data have;
set have;
by subject;
gap=trt_changeday-studyday;

retain found ;
if first.subject then found=.;

if not found and abs(gap) le 30 and gap lt 0 and lag(gap) gt 0 then do;
 found=1;flag=1;
end;

if not found and abs(gap) gt 30 then do;
 found=1;flag=1;
end;

drop found;
run;

[chuck33]

Thanks Ksharp.
I just tweaked this a little and got what I wanted,
Your insight helped.