topic Re: retain and first. statement in SAS Programming

retain and first. statement

Tibby — Tue, 20 Sep 2016 15:43:32 GMT

I've never expected the difficulty to perform a simple procuders with retain and first. statement.

My data look like this:

id	jobnum	farm_ever	startyear	endyear
1004	1	0	1955	1959
1004	2	0	1959	1962
1004	3	1	1962	1977
1008	1	0	1951	1978
1008	2	0	1978	1981
1011	1	1	1954	1998
1012	1	0	1965	1966
1012	2	1	1966	1968
1012	3	0	1968	1972
1012	4	1	1972	1975
1012	5	1	1975	2000
1014	1	0	1959	1963
1014	2	1	1963	1965
1014	3	0	1965	1967
1014	4	1	1967	1970
1014	5	1	1970	1977
1014	6	1	1978	1998
1014	7	1	1998	1999

I need to calculate duration (endyear - startyear) when farm_ever = 1. When farm_ever = 1 continuously, I need duration across all jobnum. For example, when id=1014, farm_ever = 1 at jobnum 4, 5, 6, 7, the duration should be 1999-1967. I tried to use retain and first. statement :

proc sort data = jobhist ; by id jobnum farm_ever ; run ;

data jobhist ;
set jobhist ;
by id jobnum farm_ever ;
retain start ;
if first.farm_ever then start = startyear ;
run ;

But the results did not get retained because first.farm_ever is 1 for every entry. Then I tries proc sort without jobnum, but when farm_ever = 0, it went up to first, then the wrong year got retained.

Could anyone please help me out? Is there any function to fix this, or is there any other statement to do this apart from retain?

Thank you so much!

Re: retain and first. statement

Astounding — Tue, 20 Sep 2016 16:05:55 GMT

The sorted order is correct. RETAIN is also correct. Here's how you might add to those pieces:

data want;

set jobhist;

by id farm_ever notsorted;

retain start;

if farm_ever=1 then do;

if first.farm_ever then start = startyear;

if last.farm_ever then duration = endyear - start;

end;

drop start;

run;

NOTSORTED permits the BY statement, even though the data are not in order by FARM_EVER. You can see how the program utilizes first. and last. from that point.

Re: retain and first. statement

ballardw — Tue, 20 Sep 2016 16:15:21 GMT

What about ID 1008 where farm_ever is never 1?

If the LAST value for Farm_ever going to be equal to 0 when it has been 1 in earlier jobnum values?

This may get close to the duration depending on responses to those questions.

data want ;
     set jobhist ;
     by id jobnum farm_ever ;
     retain start ;
     if first.id then start= -999;
     if farm_ever=1 and start=-999 then start=startyear;
     else if farm_ever=0 then start=-999;
     if last.id and start ne -999 then duration = endyear-start; 
run ;

BTW I recommend extreme caution with the

Data Jobhist;

set Jobhist;

code. It is entirely too easy to end up with bad data if you recode any of your original variables. Multiple passes can recode multiple times.

Re: retain and first. statement

Tibby — Tue, 20 Sep 2016 17:04:50 GMT

Thank you so much for your help and for explaining notsorted. It worked just as the way I wanted.

Re: retain and first. statement

Tibby — Tue, 20 Sep 2016 17:09:28 GMT

Thank you for your suggestion about data jobhist; set jobhist; I'll be more careful of using it.

If for ID 1008 that farm_ever has never been 1, then no need to deal with it. The tricky part is to that LAST value for Farm_ever going to be equal to 0 when it has been 1 in earlier jobnum values.

The code you provided did pretty much most of the calculation but some tricky ones were missed. It was pretty close, though. Thank you for sharing your insight!