programming for extracting eligible subjects from counting process format.

[asinusdk]

I wonder what is the general rule for extracting eligible subjects in a counting process format in SAS.

For example, take a look at the table below. Let's assume that I set a condition that a subject (ID) is not eligible if psyc_c is both 0 during 2002 and 2003. Then I wouldn't want to include the subject in the dataset. I think that I deal with a problem like this very often in programming in a counting process format. Another example can be that I want to eliminate subjects if there is a certain number in a variable in a certain year. For example, I want to eliminate a subject if their psyc_c is 3 in 2004, then ID 3 will be excluded from the data. Please help me to figure out.. Thanks.. 

Year	Id	psyc_c	dprs_c
2002	1	0	0
2003	1	0	1
2004	1	0	0
2005	1	0	0
2006	1	0	0
2002	2	0	0
2003	2	0	0
2004	2	0	0
2005	2	0	0
2006	2	0	0
2002	3	0	2
2003	3	0	0
2004	3	3	0
2005	3	0	0
2006	3	0	0
2002	4	1	0
2003	4	0	0
2004	4	0	0
2005	4	0	3
2006	4	0	3

[novinosrin]

---

@asinusdk wrote:

I wonder what is the general rule for extracting eligible subjects in a counting process format in SAS.

For example, take a look at the table below. Let's assume that I set a condition that a subject (ID) is not eligible if psyc_c is both 0 during 2002 and 2003. Then I wouldn't want to include the subject in the dataset. I think that I deal with a problem like this very often in programming in a counting process format. Another example can be that I want to eliminate subjects if there is a certain number in a variable in a certain year. For example, I want to eliminate a subject if their psyc_c is 3 in 2004, then ID 3 will be excluded from the data. Please help me to figure out.. Thanks.. 

Year	Id	psyc_c	dprs_c
2002	1	0	0
2003	1	0	1
2004	1	0	0
2005	1	0	0
2006	1	0	0
2002	2	0	0
2003	2	0	0
2004	2	0	0
2005	2	0	0
2006	2	0	0
2002	3	0	2
2003	3	0	0
2004	3	3	0
2005	3	0	0
2006	3	0	0
2002	4	1	0
2003	4	0	0
2004	4	0	0
2005	4	0	3
2006	4	0	3

---

The logical approach is pretty straight forward-->

You just need to structure your piecemeal like

1. check for condition 

2. flag if true/false accordingly

3. update/delete/or just retain the flagged group/records

4. Outout to one or many datasets based on the flagged condition if any

5. Translate the above 4 into SAS language

And you are done

[PGStats · Accepted solution]

The best strategy depends a lot on the type of rules that you want to implement. For the example given, a flexible approach would be to transpose your data and remerge the selected ids:

data have;
input Year Id psyc_c dprs_c;
datalines;
2002 1 0 0 
2003 1 0 1 
2004 1 0 0 
2005 1 0 0 
2006 1 0 0 
2002 2 0 0 
2003 2 0 0 
2004 2 0 0 
2005 2 0 0 
2006 2 0 0 
2002 3 0 2 
2003 3 0 0 
2004 3 3 0 
2005 3 0 0 
2006 3 0 0 
2002 4 1 0 
2003 4 0 0 
2004 4 0 0 
2005 4 0 3 
2006 4 0 3 
;

proc transpose data=have out=full(drop=_name_) prefix=psyc_c_;
by id;
id year;
var psyc_c;
run;

proc print data=full; run;

data selected(keep=id);
set full;
if psyc_c_2002 = 0 and psyc_c_2003 = 0 then delete;
if psyc_c_2004 = 3 then delete;
/* .... */
run;

data want;
merge have selected(in=ok);
by id;
if ok;
run;

proc print data=want; run;

[asinusdk]

I thought about transposing it, but I couldn't handle a problem some subjects who are missing for some years.... 

[PGStats]

Some subjects missing for some years wouldn't be a problem, unless they are missing for all years. Experiment!

[Peter_C]

When data are already ordered, rather than re-merge or transpose I prefer a parallel pass in (this case) ID by- group. The first pass collects the criteria that are applied on the second pass.

Data selected ;
Set have(in= a) have(in=b) ;
By ID ;
If first.id then do ;
Retain flag1-flag3 counter1-counter3 ;
Call missing( of flag1-flag3, of counter1-counter3).;
IF A then do ;
" update counters if appropriate ;
* check if criteria apply and set flags ;
End ;

If ( criteria are right ) then output ;
Run;