Selecting certain rows for new data set

[Dave13]

Please feel free to tell me to RTFM (just give friendly location). I am using SAS v9.4. I have a reasonably large data set, that I need to select certain rows to output to a new data set.I'm attaching a couple of examples. The variables are: Obs Grpelem Grindex Index Pfqmn Ffqmn Minfqmn Maxfqmn Bwmn Pkfqme Ffqme Minfqme Maxfwme Bwme Fmcthresh Sub Index1

1 2 23 1 39000 39000 38000 41000 2900 38600 38600 38400 39800 2200 . 8073F -2
2 2 25 1 70300 70300 64400 72200 7800 68100 68100 67500 69700 2600 . 8073F -2
3 2 25 2 44900 44900 43900 46800 2900 44800 44800 44000 46400 2600 23300 8073F -1

.

.

.17 2 148 1 38000 38000 36100 41000 4800 37900 37900 37900 39000 2100 . 8073F -2
18 2 178 1 37100 37100 35100 39000 3900 36200 36200 36100 37400 2300 . 8073F -2
19 4 183 1 37100 37100 35100 39000 3900 36800 36900 36600 37900 2200 . 8073F -2
20 3 201 1 36100 36100 35100 39000 3900 35900 35900 36200 36900 2000 . 8073F -2
21 3 201 2 67300 33600 66400 69300 2900 67300 44900 66400 69300 2900 9000 8073F -1
22 3 201 3 33200 33200 32200 35100 2900 33200 33200 32500 35000 2500 11700 8073F -1

I need to select the rows where Grindex has the same value in multiple rows (e.g. rows 2,3 + rows 20, 21, 22 in the example above). The actual number in the Grindex column will vary, but some rows will have repeats for each subject. These are the rows I need to put into a new data set for further processing. I can't provide any code, as this one has me stumped. The attempts I've made have failed as I can't figure out how to select multiple rows on a variable that will change multiple times within a subject and between subjects.

[unison]

Simplifying your problem, see if something like this works for your situation:

data have;
input obs grp num;
datalines;
1 1 25
2 1 25
3 1 44
4 2 3
5 3 64
6 3 64
;
run;

proc sql;
create table want(drop=_cnt) as
select *, sum(1) as _cnt from have 
group by grp,num
having _cnt gt 1
order by obs
;quit;

-unison

[Dave13]

Thanks for the response. I had looked into web searches, and suspected SQL might be needed. Will need to purchase SQL manual and read up. The code posted by another person worked, so for now I can move forward with the project. Will clearly need to learn SQL, as I am sure this type of situation will come up again with this type of data. Greatly appreciate your time for the response.

[Astounding · Accepted solution]

It's difficult to see where SUBJECT comes from or how it might be used, but this should be a good set of tools to apply:

data want;
   set have;
   by grindex notsorted;
   if first.grindex=0 or last.grindex=0;
run;

It's conceivable (but not clear) that you would need to add another variable to the BY statement.

It's 100% forgivable that you didn't find this in the documentation.  It's a combination of tools that you wouldn't normally see in your first few months of programming.

[Dave13]

This did exactly what I wanted. Thanks! Can see I need to play with this code to fully understand it, but at least I can move forward with the rest of the data cleanup/checking.

[unison]

For learning purposes, it's a good idea to play with a small sample of your data so that you're not bogged down by the details. In your case, you can take a look at what "first.grindex" and "last.grindex" are for each observation by doing something like the following (using a simplified 'have' dataset):

data have;
input obs grp num;
datalines;
1 1 25
2 1 25
3 1 44
3 2 25
4 2 3
5 3 64
6 3 64
;
run;

data want;
   set have;
   by num notsorted;
   first_val = first.num;
   last_val = last.num;
/*   if first.num=0 or last.num=0;*/
run;

Cheers!

-unison

[Dave13]

Yup, this is what I had in mind also. Just have to get existing data pulled together and cleaned up, so analysis can begin. Will have to circle back to this next week. Thanks Again!!