PROC SQL Most common variables

[takahasm]

Hello,

I want to know the top 3 most common reason for visit for each birth group for a dataset with 2 columns where:

• Column 1 has birth group (so like child, senior, adult, etc) [BD_CLASS] 
  
• Column 2 has that’s their reason for visiting the ER. [main_problem]

I tried using 

PROC SQL;
SELECT BD_CLASS, count(unique(main_problem)) as u_PROB
FROM work.test12
group by BD_Class;
quit;

but that only told me the number of unique problems in each birth group; not the most popular ones. 

Any help would be much appreciated

[ballardw]

Does this have to be done with SQL?

[takahasm]

no it doesn't! 

[Reeza]

1. Try ordering the data, so sort by U_PROB descending so that the top are on top.
2. Add OUTOBS=3 to limit to the top 3 but then you may lose any duplicates.