topic Re: Calculating duration from a Panel Structure in SAS Programming

Calculating duration from a Panel Structure

Fletcher — Wed, 17 Oct 2012 22:54:58 GMT

Hi All,

New SAS user here. I have a panel data structure, from which I am trying to calculate the duration between observations. My test data looks like this:

ID Enter Date Exit Date Sex Age

1 1/1/2000 1/1/2002 m 24

1 5/5/2006 5/5/2009 m 30

1 8/8/2011 8/8/2013 m 35

2 2/2/2002 2/2/2005 f 42

2 6/6/2006 6/6/2010 f 46

3 3/3/2003 3/3/2004 m 19

3 4/4/2008 4/4/2012 m 24

I want to create a variable that subtracts the exit date from the next enter date, so that last observation for each ID group is missing. Any efficient ways to do this without running loop?

Thanks in Advance.

Re: Calculating duration from a Panel Structure

PGStats — Wed, 17 Oct 2012 23:47:58 GMT

Don't know what you mean by "without running loop", but this is efficient :

data test;
informat enterDate exitDate mmddyy9.;
format enterDate exitDate mmddyy10.;
input ID     EnterDate     ExitDate     Sex $    Age;
datalines;
1        1/1/2000       1/1/2002       m       24
1        5/5/2006        5/5/2009      m       30
1        8/8/2011        8/8/2013      m       35
2        2/2/2002        2/2/2005       f        42
2        6/6/2006        6/6/2010       f        46
3        3/3/2003        3/3/2004      m       19
3        4/4/2008        4/4/2012      m       24
;

data want;
set test end=end;
if not end then do;
      set test (firstobs=2 keep=id enterDate rename=(enterDate=nextDate id=nextId));
      if id=nextId then duration = intck("DAY", exitDate, nextDate);
      end;
run;

proc print; run;

Re: Calculating duration from a Panel Structure

Linlin — Thu, 18 Oct 2012 01:02:26 GMT

data want;

merge test(in=a) test(firstobs=2 keep=id enterdate rename=(id=_id enterdate=_enterdate));

duration=ifn(id=_id,intck("DAY",exitdate, _enterdate),.);

if a;

drop _:;

proc print;run;

Re: Calculating duration from a Panel Structure

Haikuo — Thu, 18 Oct 2012 02:30:22 GMT

Here is an obnoxious SQL approach with the undocumented feature:

data have;

input ID (Enter_Date Exit_Date) (:mmddyy10.) Sex$ Age;

format Enter_Date Exit_Date mmddyy10.;

cards;

1 1/1/2000 1/1/2002 m 24

1 5/5/2006 5/5/2009 m 30

1 8/8/2011 8/8/2013 m 35

2 2/2/2002 2/2/2005 f 42

2 6/6/2006 6/6/2010 f 46

3 3/3/2003 3/3/2004 m 19

3 4/4/2008 4/4/2012 m 24

;

proc sql;

select a.*,b.enter_date-a.exit_date as new_date from

(select *, monotonic() as obs from have) a

left join

(select *, monotonic() as obs from have) b

on a.id=b.id and

b.obs-a.obs=1

order by id, enter_date;

quit;

Haikuo