Proc SQL: SUM multiple variables with the same prefix

[UDcc123]

I'm trying to sum multiple variables within proc sql with the same prefix without having to list them all out.  Sample is below.  Looking for code so I don't have to write out "stat1,stat2,stat3" (my real code has stat1-stat30)

data sample;

  stat1 = 10;

  stat2 = 4;

  stat3 = 1;

run;

proc sql;

  select

    SUM(stat1,stat2,stat3) as sumEQ15

    ,SUM(stat1-stat3) as sumEQ9

    ,SUM(stat1--stat3) as sumEQ11

    ,SUM(stat:) as sumEQerror /* get error with this logic */

  from sample;

quit;

[Reeza]

Unfortunately PROC SQL does not support variable lists.

Since your summation is the values in a row, you can consider a data step instead.

data want;

set sample;

sumeq15=sum(stat1, stat2, stat3);

sumeq9=sum(of stat1--stat3);

sumeqerror=sum(of stat:);

run;

[UDcc123 · Accepted solution]

@Reeza wrote:

Unfortunately PROC SQL does not support variable lists.

Since your summation is the values in a row, you can consider a data step instead.

data want;
set sample;
sumeq15=sum(stat1, stat2, stat3);
sumeq9=sum(of stat1--stat3);
sumeqerror=sum(of stat;
run;

ok...thanks.  I worked around it with a macro (see below) so I didn't need to add an extra variable in the data set, but was hoping there was a simpler solution

%macro sum1;
%do n=1 %to 3;
,stat&n
%end;
%mend sum1;
proc sql;
select
     SUM(0 %sum1) as SumByRow
from sample;
select
     SUM(SUM(0 %sum1) as SumAcrossAllRows
  from sample;
quit;

And @PGStats wrote:

As an alternative, you can transpose first and get closer to a mormalized data structure, and then use SQL :

data sample;
  id = 1;
  stat1 = 10;
  stat2 = 4;
  stat3 = 1;
run;
 
proc transpose data=sample out=sampleList;
by id;
var stat:; /* Or stat1 - stat3, or stat1 -- stat3 */
run;
 
proc sql;
create table want as
select id, sum(col1) as sumEQ
from sampleList
group by id;
quit;
 
proc print data=want noobs; run;

[Reeza]

It doesn't work for me, but if it works for you that's good

I would recommend testing your code with multiple lines of data to ensure you obtain the expected results:

data sample;

  stat1 = 10;

  stat2 = 4;

  stat3 = 1;

  output;

  stat1 = 20;

  stat2 = 8;

  stat3 = 2;

  output;

run;

data want;

set sample;

sumeq15=sum(stat1, stat2, stat3);

sumeq9=sum(of stat1--stat3);

sumeqerror=sum(of stat:);

run;

[UDcc123]

it works with multiple lines of code depending on what you want.

%macro sum1;

%do n=1 %to 3;

,stat&n

%end;

%mend sum1;

proc sql;

select

     SUM(0 %sum1) as SumByRow

from sample;

select

     SUM(SUM(0 %sum1) as SumAcrossAllRows

  from sample;

quit;

[PGStats]

As an alternative, you can transpose first and get closer to a mormalized data structure, and then use SQL :

data sample;

  id = 1;

  stat1 = 10;

  stat2 = 4;

  stat3 = 1;

run;

proc transpose data=sample out=sampleList;

by id;

var stat:; /* Or stat1 - stat3, or stat1 -- stat3 */

run;

proc sql;

create table want as

select id, sum(col1) as sumEQ

from sampleList

group by id;

quit;

proc print data=want noobs; run;

PG