How to select specific records

[hjjijkkl]

I am trying to keep records with results <=50. If one the ID has results >50 I want to remove  the subject from further analysis. How can I do that?

ID	Result
1	50
1	23
2	16
2	7
2	26
3	55
3	30
3	43
3	5
4	50
4	23

I want an output like this

ID	Result
1	50
1	23
2	16
2	7
2	26
4	50
4	23

[Reeza]

proc sql;
create table want as
select t1.* from have t1
where ID not In (select distinct t2.ID from have t2 where t2.results > 50);
quit;

---

@hjjijkkl wrote:

I am trying to keep records with results <=50. If one the ID has results >50 I want to remove  the subject from further analysis. How can I do that?

ID	Result
1	50
1	23
2	16
2	7
2	26
3	55
3	30
3	43
3	5
4	50
4	23

I want an output like this

ID	Result
1	50
1	23
2	16
2	7
2	26
4	50
4	23

---

[hjjijkkl]

is there another way to do it besides proc sql?

[Reeza]

Some other methodologies include:

• Sort by ID and descending result, so that the first record per ID s the largest. If that's greater than 50 then you need to drop that record otherwise you can keep it.
• Create a list of all IDs with a result > 50. Then merge the two tables together, using the data set IN option to select only records that are not in the merged table. 
• DoW loops - fairly complex but allow you to summarize data ahead of time and then merge the results in. Use similar logic as the first approach

All of these essentially have two passes of the data in one form or another but SQL and the DoW loop let it "appear" to be one step.

---

@hjjijkkl wrote:
is there another way to do it besides proc sql?

---

[Astounding · Accepted solution]

FWIW, this appears to be closer to what you are looking for.  However, it can give you a note in the log about more than one data set have repeats of the BY variable:

data want;
   merge have have (where=(result > 50) in=delete_me);
   by id;
   if delete_me then delete;
run;