topic Re: Not sure how to summarize my question... in SAS Programming

Not sure how to summarize my question...

mcook — Mon, 08 Jun 2020 19:52:04 GMT

I am trying to create 4 new variables(var_A, Var_B, Var_C, Var_D). Where Var_A has a value of 1 if Response is A, and a value of 0 otherwise. And Var_B has a value of 1 if Response is B , and a value of 0 otherwise. Likewise for C and D.

I have done it the following 2 ways, and am wondering if there is a built in function for such a process.

** Way 1:

Data Table1;
set Table1;

if Response = "A" then
Var_A = 1;
else if Response = "B" then
Var_B = 1;
else if Response = "C" then
Var_C = 1;
else if Response = "D" then
Var_D = 1;
run;

/*macros to add zeros to the non 1 places from above*/
%LET List= A B C D;

%Macro Zeros();

data Table1;
set Table1;

if Var_&MM. ^= 1 then
Var_&MM. = 0;
run;

%MEND Zeros;

%Macro Loop_M();
%DO M=1 %TO 4;
%LET MM=%SCAN(&List., &M.);

%Zeros();
%END;
%MEND Loop_M;

%Var_M();

** Way 2:

Data Table1;
set Table;

if Response = "A" then do
Var_A=1; Var_B=0; Var_C=0; Var_D=0;
end;
else if Response = "B" then do
Var_B=1; Var_A=0; Var_C=0; Var_D=0;
end;
else if Response = "C" then do
Var_C=1; Var_A=0; Var_B=0; Var_D=0;
end;
else if Response = "D" then do
Var_D=1; Var_A=0; Var_B=0; Var_C=0;
end;
else do
Var_A=0; Var_B=0; Var_C=0; Var_D=0;
end;
run;

Is there an easier way to accomplish this?

Thanks

Re: Not sure how to summarize my question...

smantha — Mon, 08 Jun 2020 19:59:14 GMT

Data Table1;
     set Table;
          Var_A=0; Var_B=0; Var_C=0; Var_D=0;
         if Response = "A" then do
              Var_A=1;
        end;
        else if Response = "B" then do
              Var_B=1;
        end;
        else if Response = "C" then do
              Var_C=1; 
        end;
        else if Response = "D" then do
              Var_D=1; 
        end;
       
run;

Data Table1;
     set Table;
          
        Var_A = 1* (Response='A');
         Var_B = 1* (Response='B');
         Var_C = 1* (Response='C');
           Var_D = 1* (Response='D');
     
        
       
run;

Data Table1;
     set Table;
        %let i = 1 %to 4;
            %let varlst = A B C D;
            %let var=%scan(&varlst.,&i.);
              Var_&var.= 1* (Response="&var."); 
        %end;  
             
run;

Three ways hown above

Re: Not sure how to summarize my question...

s_lassen — Mon, 08 Jun 2020 20:04:38 GMT

It could be done with arrays, e.g.:

data want;
  set have;
  array responses(4) $1 _temporary_ ('A','B','C','D');
  array vars(4) 8 Var_A Var_B Var_C Var_D;
  do _N_=1 to dim(vars);
    vars(_N_)=(Response=responses(_N_));
    end;
run;

Re: Not sure how to summarize my question...

PaigeMiller — Mon, 08 Jun 2020 20:10:16 GMT

SAS has built-in ways to create 0/1 (sometimes called "dummy" variables) so you don't have to program it yourself (and risk getting it wrong). One method is PROC GLMMOD, but there are many other methods as discussed in this thread How to create dummy variables - Categorical Variables

In addition, many SAS procedures allow you to use a CLASS statement, so the 0/1 variables are created by the PROC internally, so you don't have to create them at all. Please consider this approach, as it is the least painful and least error-prone method.

But I guess the real question is, what are you going to do with these 0/1 variables once you create them?

Re: Not sure how to summarize my question...

ballardw — Mon, 08 Jun 2020 20:13:16 GMT

A better question may be "Why do you need 4 variables?".

One of the typical uses of what you are requesting is to use the 1/0 coded values as "dummy" or "indicator" variables in regression analysis. Most of the SAS regression procedures will allow you to place your Response variable on a CLASS statement so that when it is used in the Model the levels are created as needed.

And another way

proc format;
invalue AA
'A'=1
other=0
;
invalue BB
'B'=1
other=0
;
invalue CC
'C'=1
other=0
;
invalue DD
'D'=1
other=0
;
run;

data example;
   input response $;
   a = input(response,aa.);
   b = input(response,bb.);
   c = input(response,cc.);
   d = input(response,dd.);
datalines;
A
D
B
C
;

The informats have a 2 character name because a single D will conflict with a SAS supplied informat, or at least that was the error I got. So make them all similar in name with 2 characters.

Re: Not sure how to summarize my question...

smantha — Mon, 08 Jun 2020 20:15:49 GMT

I believe you didn't intent to use _n_;

Re: Not sure how to summarize my question...

Ksharp — Tue, 09 Jun 2020 12:00:34 GMT

data have;
input response $;
cards;
A
B
C
D
;
data temp(rename=(response=var_));
 set have;
 y=1;
run;
proc logistic data=temp outdesign=design(drop=y) outdesignonly;
class var_ / param=glm ;
model y=var_/nofit noint;
run;
data want;
 merge have design;
run;

Re: Not sure how to summarize my question...

mcook — Tue, 09 Jun 2020 17:58:19 GMT

I really liked that 2nd way. I never thought of doing it like that.