Combining Variables

Sbon08 — Fri, 08 Mar 2019 18:54:08 GMT

Hi everyone, my dataset in excel has multiple columns for race. Each column has a 1 if someone chose that race and a blank they did not (they would have a 1 in another column). I am trying to combine these all into one race variable for a logistic regression. Attached is a picture of how the raw data looks:

Re: Combining Variables

ballardw — Fri, 08 Mar 2019 21:51:10 GMT

@Sbon08 wrote:

Hi everyone, my dataset in excel has multiple columns for race. Each column has a 1 if someone chose that race and a blank they did not (they would have a 1 in another column). I am trying to combine these all into one race variable for a logistic regression. Attached is a picture of how the raw data looks:

Is this going to be a dependent variable or independent in the model?

What would the values of your one value look like? Are all of the Q2_1 through Q2_9 to be combined or does one of those mean something such that it should not be combined such as a "refused to answer" code that really should not occur with other values?

And are the variables numeric or character?

By any chance did a previous step replace zero (0) values with missing? If so remove that step and something like this would work:

data example;
   input Q2_1 - Q2_4;
   length r $ 4;
   r = cats(of Q2_:);

datalines;
1 0 0 0
1 0 1 0
0 0 0 1
; 
run;

Using only 4 variables as I'm too lazy to make 10 when 4 demonstrates what goes on.

You would need a place holder in any case.

The following code shows one way to replace 0 for missing values.

data example;
   informat Q2_1 - Q2_4 $1.;
   input Q2_1 - Q2_4 ;
   array q  Q2_1 - Q2_4;
   do i= 1 to dim(q);
      if missing(q[i]) then q[i]='0';
   end;
   length r $ 4;
   r = cats(of Q2_:);
   drop;
datalines;
1 . . .
1 . 1 .
. . . 1
; 
run;

If you variables are numeric the only real change to the code (other than not reading data) is to use =0; instead of ='0';

topic Re: Combining Variables in SAS Programming

Combining Variables

Re: Combining Variables