topic Re: Flagging sourse with PROC SQL in SAS Programming

Flagging sourse with PROC SQL

BayzidurRahman — Mon, 17 Jul 2023 02:50:02 GMT

How can we flag the source of each observations after merging two datasets using PROC SQL? The flag variable should identify whether the observation belongs to both datasets or any of them.

Re: Flagging sourse with PROC SQL

SASKiwi — Mon, 17 Jul 2023 03:03:15 GMT

Use CASE logic:

proc sql;
  select case
            when not missing(A.column) and not missing(B.column) then 'BOTH'
            when not missing(A.column) then 'FROMA'
            when not missing(B.column) then 'FROMB'
            else ' '
         end as Source_Flag length = 5
  from table1 as A
  outer join table2 as B
....;
quit;

Re: Flagging sourse with PROC SQL

Tom — Mon, 17 Jul 2023 03:18:28 GMT

The easiest way is to use normal SAS code to merge instead since it actually has a concept of source. So code like this will create a FLAG variable with values of 1, 2 or 3 (both).

data want;
  merge a(in=in1) b(in=in2);
  flag = 2*in2 + in1 ;
run;

You could do something similar in SQL code with a little work.

Say you had this SQL code:

proc sql;
create table want as
  select ...variable list...
  from A 
  full join B
  on (...criteria...)
;

You could add those IN1 and IN2 flags and the create the new FLAG variable the same way like this:

create table want as
  select ...variable list...
    ,case when (in1 and in2) then 3 when (in1) then 1 else 2 end as FLAG
  from (select *,1 as in1 from A ) A
  full join (select *,1 as in2 from B) B
  on (...criteria...)
;

You need the CASE instead of the simple arithmetic because the IN1 and IN2 variables will be coded as 1 and missing instead of the 1 and 0 that they would have had when created by the IN= dataset option.

Re: Flagging sourse with PROC SQL

BayzidurRahman — Mon, 17 Jul 2023 05:47:11 GMT

Thanks for the code.

When I ran the last set of code with PRC SQL, the flag variable is created ok but the matching variable is missing when flag=2.

Re: Flagging sourse with PROC SQL

Kurt_Bremser — Mon, 17 Jul 2023 07:27:19 GMT

You need to use the COALESCE function for your key variable(s) in the SELECT.

select
  coalesce(a.key,b.key) as key,
  .....

Re: Flagging sourse with PROC SQL

BayzidurRahman — Mon, 17 Jul 2023 08:37:06 GMT

The combined dataset now contains only the key variable.