topic Re: Creating dummy variable in SAS Procedures

Creating dummy variable

kthartma — Sun, 18 Mar 2018 23:06:10 GMT

Hi there,

I want to create a dummy variable for gender (m=0, f=1) so that I can use proc reg for a bivariate analyses. However, my error keeps indicating that my variable doesn't exist. Any insight would be useful 🙂

data eva.cohort;
set eva.finalcohort;

if sex="m" then sexm=0;

if sex="f" then sexf=1;
avgvol1=sum(avg_2009+avg_2011+avg_2012+avg_2014)/4;
avgvol2=sum(bee_avg_2009+bee_avg_2011+bee_avg_2012+bee_avg_2014)/4;
run;

proc means data=eva.cohort n nmiss mean median max min;

var avgvol1 avgvol2;

run;

proc reg data=eva.cohort;

model avgvol1=sexm sexf;

run;

Re: Creating dummy variable

Astounding — Sun, 18 Mar 2018 23:23:58 GMT

Perhaps your actual data doesn't contain "m" and "f". Perhaps it contains "M" and "F" instead.

At any rate, you would be well advised to treat SEX as a CLASS variable within PROC REG. Most regression procedures will automatically create the proper dummy variables when you use a CLASS statement.

Note for the future: instead of posting the program, post the log so we can see what message applies to what step.

Re: Creating dummy variable

PaigeMiller — Mon, 19 Mar 2018 00:46:02 GMT

@Astounding wrote:

At any rate, you would be well advised to treat SEX as a CLASS variable within PROC REG. Most regression procedures will automatically create the proper dummy variables when you use a CLASS statement.

PROC GLM, not PROC REG

To @kthartma: there is no need to create your own dummy variables. PROC GLM will create them for you, and also avoid the programming error you are having.

Re: Creating dummy variable

Reeza — Mon, 19 Mar 2018 00:50:02 GMT

You're not using the SUM() function as usually intended either.

It's usually used when you want to consider missing as 0, this approach wouldn't do that because you've listed the items with + in between rather than comma's.

Test your code with the following:

avgvol1=sum(avg_2009+avg_2011+avg_2012+avg_2014)/4;

avgvol1_check0 =sum(avg_2009, avg_2011, avg_2012, avg_2014)/4;
avgvol1_check1 = sum(of avg_2009-avg_2014)/ 4;
avgvol1_check2 = mean(of avg_2009-avg_2014);

If you have different results between any of the calculations you have an issue.

Re: Creating dummy variable

PGStats — Mon, 19 Mar 2018 03:45:23 GMT

If you insist on using proc reg your code should read:

data eva.cohort;
set eva.finalcohort;

if sex="m" then sexDum=0;
if sex="f" then sexDum=1;

avgvol1 = (avg_2009 + avg_2011 + avg_2012 + avg_2014) / 4;
avgvol2 = (bee_avg_2009 + bee_avg_2011 + bee_avg_2012 + bee_avg_2014) / 4;
run;

proc means data=eva.cohort n nmiss mean median max min;
var avgvol1 avgvol2;
run;

proc reg data=eva.cohort;
model avgvol1 = sexDum;
run;

Re: Creating dummy variable

kthartma — Sat, 24 Mar 2018 17:45:46 GMT

Thanks so much! I ended up using proc glm.