I would like to do the calculation only the variable exists within a dataset

[Sami1234]

Wondering if someone can help me, please?

I would like to create another variable depending on if the variable exists in the dataset.

If the variable does not exists in the dataset then no need to create the variable.

At the moment, it's creating a blank variable for M0 and STD0 (as they don't exist) and throwing a warning in the log for the uninitialized variables.

Also, would it be possible to create a loop or an array to do the checks and create (diff0-diff9) variables for up to 10 variables (m0-m9) depending on if the correspondence variables exist?

Thank you!

data testb;

set testa;

if _n_=1 then do;

dsid=open('testa');

varexistm0 = varnum(dsid,'M0') ;

varexistm1 = varnum(dsid,'M1') ;

varexistm2 = varnum(dsid,'M2') ;

varexists0 = varnum(dsid,'STD0') ;

varexists1 = varnum(dsid,'STD1') ;

varexists2 = varnum(dsid,'STD2') ;

rc=close(dsid);

end;

drop rc dsid;

if varexistm0 > 0 and  varexists0 >0 then do;

diff0= (m10 - m0) / sqrt((std10**2 + std0**2) / 2);      

end;

if varexistm1 > 0 and  varexists1 >0 then do; then do;

diff1= (m10 - m1) / sqrt((std10**2 + std1**2) / 2);

 end;      

if varexistm2 > 0 and  varexists2 >0 then do;

diff2= (m10 - m2) / sqrt((std10**2 + std2**2) / 2);

end;

run;

[Kurt_Bremser]

Pull the metadata from dictionary.columns, and then use da data _null_ step to dynamically generate the code:

(I used a simple subtraction as the mathematical operation to be performed)

data have;
input id $ m0 m1 std0 std1;
datalines;
x 1 2 3 5
;
run;

proc sql noprint;
create table variables as
select name, substr(name,length(name)) as seq
from dictionary.columns
where libname = 'WORK' and memname = 'HAVE' and anydigit(calculated seq);
select name into :vars separated by ' ' from variables order by seq, name;
quit;

data _null_;
call execute("data want; set have;");
do i = 1 to countw("&vars.") by 2;
  put _all_;
  targetvar = "diff" !! substr(scan("&vars.",i),length(scan("&vars.",i)));
  call execute(targetvar !! '=' !! scan("&vars.",i+1) !! '-' !! scan("&vars.",i) !! ';');
end;
call execute("run;");
run;

proc print data=want noobs;
run;

Result:

id    m0    m1    std0    std1    diff0    diff1

x      1     2      3       5       2        3  

[s_lassen · Accepted solution]

The variables are created even if you do not write to them in your datastep.

I would split the code in two datasteps, and use the first datastep to generate code for the second:

filename tempsas temp;

data _null_;
dsid=open('testa');
varexistm0 = varnum(dsid,'M0') ;
varexistm1 = varnum(dsid,'M1') ;
varexistm2 = varnum(dsid,'M2') ;
varexists0 = varnum(dsid,'STD0') ;
varexists1 = varnum(dsid,'STD1') ;
varexists2 = varnum(dsid,'STD2') ;
rc=close(dsid);
end;
 
file tempsas;
if varexistm0 > 0 and  varexists0 >0 then 
  put 'diff0= (m10 - m0) / sqrt((std10**2 + std0**2) / 2);';      
 
if varexistm1 > 0 and  varexists1 >0 then 
  put 'diff1= (m10 - m1) / sqrt((std10**2 + std1**2) / 2);';

if varexistm2 > 0 and  varexists2 >0 then 
put 'diff2= (m10 - m2) / sqrt((std10**2 + std2**2) / 2);';
run;


data testb;
  set testa;
  %include tempsas/source2;
run;
  

[Sami1234]

Thank you so much, that works absolutely fine without any warnings.