Categorical => assigns categories. In this case that would be a variable that indicates whether a person is a current/former smoker or Non-smoker. Shouldn't really think vs "general pop", the other group as to be separated based on the characteristic of interest.
Since you are asking about the mean cholesterol then likeliest is a T-test. The input data would expect to use two variables, one the category of smoker/ non-smoker and the other the measured cholesterol.
The basic approach would be something like:
Proc Ttest data =have;
class smokerstatus;
var cholestorol;
run;
The class variable can have only two values (or have a format apply that creates exactly two levels). Any record missing the class variable will be excluded.
The Ttest expects the data to have equal variance. If you have evidence, such as from running Proc Univariate on your data with similar grouping, that the variances are not the same between groups then you would adjust by using the COCHRAN option on the proc statement to tell the proc to use a different comparison calculation.
