Looping to recode variables in a SAS dataset

07 July 2012 9 4K Report

Does anyone have a recommendation for how to loop through a series of variables in a SAS dataset which all need to be recoded the same way? For example, I am doing the following to recode the variable CF1:

DATA COGFLEX; SET COGFLEX;

SELECT(CF1);

WHEN ('1 Strongly Disagree') CF1=1;

WHEN ('2 Disagree') CF1=2;

WHEN ('3 Slightly Disagree') CF1=3;

WHEN ('4 Slightly Agree') CF1=4;

WHEN ('5 Agree') CF1=5;

WHEN ('6 Strongly Agree') CF1=6;

OTHERWISE CF1=.;

END;

RUN;

I want to do the same for variables CF2 through CF12. I am assuming there is a way to tell SAS to loop through these variables and perform the exact same transformation, but I haven't been able to figure out how to do that. Any help would be much appreciated.

Nora H Ruel

you can use an array statement: Array CFvars[12] $ CF1-CF12; but you will need to create a separate array for the numeric recode. ARRAY CFrecode[12] CFr1-CFr12; Then use a 'do loop' to loop through all 12 variables. Just make sure recoded variables are from the numeric array (CFrecode) and the select statement refers to the character array (CFvars). Hopefully that's enough to get you going. If you look up the syntax for do loops you should be able to figure it out quickly.

Russ Clay

Thank you - Yes, I had been trying to use do loops, and I think I understand how that syntax works, but did not know about creating the array. I will play with this and give it a try. Thank you very much!

Guillermo Enrique Ramos

With your code, cf1 remains an alfanumeric field and the value that you get is: " 1".

You may prefer a shorter field then you will have to short it: Then the code may be:

DATA COGFLEX; SET COGFLEX;

length cfr1 $ 1;

cfr1=substr(cf1,1,1);

*the function substr(field,beginning, size) extract bytes from cf1 beginning in the first byte and a total of 1 byte;

RUN;

As you have several fields like cf1, for example cf1-cf12, you can define an array and made the same extraction for each variable.

DATA COGFLEX; SET COGFLEX;

array cf{12} cf1-cf12;

array cfr{12} ;

length cfr1-cfr12 $ 1;

do i=1 to 12;

cfr{i}=substr(cf1{i},1,1);

end;

*the function substr(field,beginning, size) extract bytes from cf1 beginning in the first byte and a total of 1 byte;

RUN;

If you want that the new variables were numeric you can use:

DATA COGFLEX; SET COGFLEX;

array cf{12} cf1-cf12;

array cfr{12} ;

do i=1 to 12;

cfr{i}=substr(cf1{i},1,1)+0;

end;

*the function substr(field,beginning, size) extract bytes from cf1 beginning in the first byte and a total of 1 byte;

RUN;

Lidia Loukine

Hi Russ,

try this:

%macro replacement (input_data=,

b=,output_data=);

data &output_data;

set &input_data;

%do i=1 %to &b;

SELECT(CF&i);

WHEN ('1 Strongly Disagree') CF&i='1';

WHEN ('2 Disagree') CF&i='2';

WHEN ('3 Slightly Disagree') CF&i='3';

WHEN ('4 Slightly Agree') CF&i='4';

WHEN ('5 Agree') CF&i='5';

WHEN ('6 Strongly Agree') CF&i='6';

OTHERWISE CF&i=' ';

END;

%end;

run;

%mend;

%replacement (input_data=COGFLEX,

b=12,output_data=COGFLEX_test);

It should work. I changed the numeric values to the character values. You cannot change the variable type as it was stated in your code sample. If you want to change "1" (character) to 1 (numeric) you should do it later.