Creating dummy variables in SAS from a categorical variable - loops

Question

I am looking to create dummy variable for a categorical variable in SAS. The categorical variable includes information on sites and takes on values such as Manila, Rabat etc., all in all there are about 50 different sites. What would be the most efficient way to create dummies without creating each dummy separately using "if then"? Maybe using loops? How would that look like

Answer 1

short answer: yes. Without further input I'm afraid there is little we can provide. here are a few examples:
data with_categoric(keep=category:);
set sashelp.zipcode;
category1 = (TIMEZONE='Central' and length(COUNTYNM) <=4);
if 35>Y then category2='low';
else if 35<Y<41 then category2='medium';
else category2='high';
run;
An alternative way to do the Category2 is via proc format:
proc format;
value level
low-35 = 'low'
35-41 = 'med'
41-high ='high';
quit;
data W_proc_foramt;
set sashelp.zipcode;
levelled = Y;
format levelled level.;
run;
Your can check more from Documentation

Answer 2

The easiest way to create the dummy category is to use the observation number as a suffix.
Solution:
/*Create Table with 5 Records*/
data input;
input Category $40.;
cards;
A
B
C
D
E
;;;;
run;
/*Create dummy categories using "_N_" record number as suffix */
data work.dummy;
set work.input;
dummy= catx("-","CAT",put(_N_,8.));
put _all_;
run;
Output:
Category=A dummy=CAT-1 _ERROR_=0 _N_=1
Category=B dummy=CAT-2 _ERROR_=0 _N_=2
Category=C dummy=CAT-3 _ERROR_=0 _N_=3
Category=D dummy=CAT-4 _ERROR_=0 _N_=4
Category=E dummy=CAT-5 _ERROR_=0 _N_=5

Answer 3

I needed to convert categorical into dummy variable in SAS and run linear regression, but did not find one place with all answers, so I will put here the result of my search.
Say we have a dataset (mydata) with dependent variable Y and categorical variables A1,A2...An. Each idependent variable has X1,X2...Xm valid values. e.g.:
A1 | A2 | A3
---|----|---
x1 | y1 | z1
x2 | y1 | z2
x1 | y2 | z3
The output after dummy conversion would be:
A1x1 | A1x2 | A2y1 | A2y2| A3z1| A3z2 | A3z3
-----|------|------|-----|-----|------|-----
1 | 0 | 1 | 0 | 1 | 0 | 0
0 | 1 | 1 | 0 | 0 | 1 | 0
1 | 0 | 0 | 0 | 0 | 0 | 1
The code to accomplish the conversion to dummy is:
DATA mydata;
set mydata;
dummy=1;
RUN;
PROC logistic data=mydata outdesignonly outdesign=design;
CLASS A1 A2 A3/param=glm;
MODEL dummy=A1 A2 A3;
RUN;
DATA mydata_dummy;
merge mydata(drop=dummy) design(drop=dummy intercept);
RUN;
DATA mydata_dummy;
SET mydata_dummy;
DROP A1 A2 A3;
RUN;
The side effect of covnerting categorical variables into dummy variables, is the inflation in varibale names.
To avoid listing all new column names (e.g. for REG)
You cannot use (MODEL Y=all) because Y will be also in all
Instead of.
MODEL Y=A1x1 A1x2 A2y1 A2y2 A3z1 A3z2 A3z3
Do the following:
PROC CONTENTS data=mydata_dummy noprint out=_contents_;
RUN;
PROC sql noprint;
SELECT name into :names separated by ' '
from _contents_ where upcase(name) ^='Y';
RUN;
PROC reg DATA=mydata_dummy;
MODEL Y=&names;
RUN;
....Thanks