can we randomly assign values to data based on id

[noda6003]

i am trying to assign val variable for the data as first id should have val=12 and second id as val=15 and the rest of id with val=20.

data test;
input pid $ date $;
cards;
1      1/1/2011   
1      1/1/2011   
1      1/3/2011 
1      3/4/2011   
2      5/1/2010  
2      6/3/2010 
;
run;

Tried something like below but it did not work.

proc sort data=test;
by pid date;
run;
 
data want;
 set test ;
 by pid ;
 if first.date then val=12;
 seq_id+3;
 if last.date then val=20;
run;

Need something like below
data need;
input pid $ date $ val ;
cards;
1 1/1/2011 12
1 1/1/2011 15
1 3/4/2011 20
2 5/1/2010 12
2 6/3/2010 15
2 6/4/2010 20
2 12/4/2010 20
2 14/4/2010 20

;
run;

Any guidance please

[RW9]

Something like:

data want (drop=seq);
  set test;
  retain seq;
  seq=ifn(first.id,1,seq+1); 
  if seq=1 then val=12;
  else if seq=2 then val=15;
  else val=20;
run;

[FreelanceReinh · Accepted solution]

Hi @noda6003,

Your PROC SORT step is good. Then you could use the LAG and LAG2 functions:

data want;
set test;
val=12+3*(pid=lag(pid))+5*(pid=lag2(pid));
run;

[noda6003]

Thanks a lot and it worked 

[Kurt_Bremser]

Do the by-group processing for the correct variable (pid), and use a simple if-then-else:

data have;
input pid $ date :ddmmyy10.;
format date ddmmyy10.;
cards;
1      1/1/2011   
1      1/1/2011   
1      1/3/2011 
1      3/4/2011   
2      5/1/2010  
2      6/3/2010
2      6/4/2010
2      12/4/2010
2      14/4/2010
;
run;

data want;
set have;
by pid;
retain val;
if first.pid then val = 12;
else if val = 12 then val = 15;
else val = 20;
run;

proc print data=want noobs;
run;

Result:

pid          date    val

 1     01/01/2011     12
 1     01/01/2011     15
 1     01/03/2011     20
 1     03/04/2011     20
 2     05/01/2010     12
 2     06/03/2010     15
 2     06/04/2010     20
 2     12/04/2010     20
 2     14/04/2010     20