topic Re: Re-catorgorize data into new variables in SAS Procedures

Re-catorgorize data into new variables

UGAstudent — Fri, 12 Jan 2018 21:22:04 GMT

I'm trying to organize a continuous variable into smaller blocks. For example. Im using CD4 counts 1-1500. I want to break the CD4 count variable into smaller blocks for analyzation. 1-200 would be low CD4. 201-500 would be medium CD4 level and 501 and above would be high CD4 level.

what code would I use to do this, if it could be done?

Re: Re-catorgorize data into new variables

novinosrin — Fri, 12 Jan 2018 21:32:01 GMT

Use proc format aka user defined format

This link might be of help-->

https://stats.idre.ucla.edu/wp-content/uploads/2016/02/bt3001.pdf

http://www2.sas.com/proceedings/sugi27/p056-27.pdf

Re: Re-catorgorize data into new variables

ballardw — Fri, 12 Jan 2018 23:10:42 GMT

@UGAstudent wrote:

I'm trying to organize a continuous variable into smaller blocks. For example. Im using CD4 counts 1-1500. I want to break the CD4 count variable into smaller blocks for analyzation. 1-200 would be low CD4. 201-500 would be medium CD4 level and 501 and above would be high CD4 level.

what code would I use to do this, if it could be done?

An example:

proc format library=work;
/* Format names cannot end in a digit to
   avoid confusion with display length options
*/
value cd4_
0  -  200 = 'Low'
200<- 500 = 'Medium'
500<-high = 'High'
;
run;

proc freq data=yourdata;
  tables cd4;
  format cd4 cd4_. ;
run;

Formats are very powerful tools as you can create multiple formats and use as needed. Almost all of the analysis procedures will honor groupings assigned in custom formats.

Re: Re-catorgorize data into new variables

UGAstudent — Fri, 12 Jan 2018 21:54:33 GMT

This worked well.

However, Im trying to do a linear regression but it says the newly created variables are not found?

proc Reg data=Import;
title "Example of linear regression";
model Factor1 = Low;
run;

Re: Re-catorgorize data into new variables

PaigeMiller — Fri, 12 Jan 2018 22:00:56 GMT

Low is not a variable. It is a formatted value.

If your goal is to do a regression on just the low values, none of the formatting is needed.

proc Reg data=Import(where=(0<=cd4<=200));
title "Example of linear regression";
model Factor1 = cd4;
run;

P.S.: It is always helpful to say what analysis you would like to do in your original post.