topic Re: how to transpose data in proc sql in SAS Programming

how to transpose data in proc sql

thanikondharish — Thu, 08 Jul 2021 07:26:18 GMT

id type score

101 testA 89

101 testB 89

102 testA 89

102 testB 89

I have one dataset like above how to transpose the data in proc sql ?

Output should be like see below example:

id testA testB

101 89 89

We should do in proc sql only.

Re: how to transpose data in proc sql

singhsahab — Thu, 08 Jul 2021 08:02:44 GMT

@thanikondharish . it's possible with the help of case and Join statement.


proc sql ;
create table want as
select a.id,testA,testB from
(select id, 
case when type eq 'testA' then score end as testA
from have where calculated testA ne .) as a,
(select id, 
case when type eq 'testB' then score end as testB
from have where calculated testb ne .) as b where a.id eq b.id;
quit;

Re: how to transpose data in proc sql

thanikondharish — Thu, 08 Jul 2021 08:03:36 GMT

I understood the code .
If we have a number of tests, what can we do.

Re: how to transpose data in proc sql

singhsahab — Thu, 08 Jul 2021 08:04:59 GMT

Better to use PROC TRANSPOSE or ARRAY.

Re: how to transpose data in proc sql

Kurt_Bremser — Thu, 08 Jul 2021 08:13:12 GMT

data have;
input id $ type $ score;
datalines;
101 testA 89
101 testB 90
102 testA 91
102 testB 92
;

proc sql;
create table want as
  select
    id,
    max(case
      when type = "testA"
      then score
      else .
    end) as testA,
    max(case
      when type = "testB"
      then score
      else .
    end) as testB
  from have
  group by id
;
quit;

But mind that this will be a maintenance nightmare if the contents of variable type change, or need some extra macro coding to make it adaptive.

That's why there must be a VERY GOOD reason to not do it with PROC TRANSPOSE, which is much simpler to code and maintains itself:

proc transpose data=have out=want (drop=_name_);
by id;
id type;
var score;
run;

Also note that wide datasets are always harder to code against, and if this is supposed to create a report, PROC REPORT does this on the fly:

proc report data=have;
column id score,type;
define id / group;
define score / "" analysis;
define type / "" across;
run;

Re: how to transpose data in proc sql

Kurt_Bremser — Thu, 08 Jul 2021 08:13:49 GMT

@thanikondharish wrote:
I understood the code .
If we have a number of tests, what can we do.

Maxim 14. Use PROC TRANSPOSE or PROC REPORT, and be done with it.

Re: how to transpose data in proc sql

PaigeMiller — Thu, 08 Jul 2021 10:20:22 GMT

@Kurt_Bremser says:

But mind that this will be a maintenance nightmare if the contents of variable type change, or need some extra macro coding to make it adaptive.

That's why there must be a VERY GOOD reason to not do it with PROC TRANSPOSE, which is much simpler to code and maintains itself:

and

Also note that wide datasets are always harder to code against, and if this is supposed to create a report, PROC REPORT does this on the fly:

also

Maxim 14. Use PROC TRANSPOSE or PROC REPORT, and be done with it.

All excellent advice.

This ought to end any discussion of doing a transpose using SQL. I hope the original poster is paying attention.

Re: how to transpose data in proc sql

ballardw — Thu, 08 Jul 2021 15:58:46 GMT

Just reading the title of this post "transpose data in proc sql" gave me a headache.

Given the verbose nature of having to name everything explicitly in SQL and the row nature of normal SQL behavior the question is sort of an extreme version of using a spoon to hammer nails. It may work but you will put a lot of effort into getting it to work in any specific case. Minor changes like adding a single value to one variable will add to the work.