Handout13 Stat3900/4950

How to take a Stratified random sample from a dataset.

source: UCLA Academic Technology Services, Stat Computing, Statistical Computing Resources

http://www.ats.ucla.edu/stat/

http://www.ats.ucla.edu/stat/sas/faq/

http://www.ats.ucla.edu/stat/sas/faq/statified_samples.htm

Using the SAS datafile hsb2.sas7bdat

------------------------------------------------------------------------------

* Using Proc SurveySelect;

proc sort data = "D:\hsb2";
by female;
run;
options nolabel;
proc means data = "D:\hsb2";
by female;
run;
* 50% sample by female;
proc surveyselect data = "D:\hsb2" out = samp1 method = srs samprate = .5 seed = 9876;
strata female;
run;
proc sort data = samp1;
by female;
run;
proc means data = samp1;
by female;
run;
* Merge the data to see which observations were selected;
proc sort data = "D:\hsb2";
by id;
run;
proc sort data = samp1;
by id;
run;
data merge1;
set "D:\hsb2" samp1;
by id;
run;
proc print data = merge1 (obs = 25);
run;
* Using more than one strata;
proc sort data = "D:\hsb2";
by female ses prog;
run;
proc means data = "D:\hsb2";
by female ses;
run;
* Similar to above with options added;
proc surveyselect data = "D:\hsb2" out = samp2 method = srs samprate = .5 seed = 9876;
strata female ses;
run;
proc sort data = samp2;
by female ses og;
run;
proc means data = samp2;
by female ses;
run;
*Using different sampling rates within each strata;
proc surveyselect data = "D:\hsb2" out = samp3 method = srs samprate = (.2 .8 .4 .1 .5) seed = 9876;
strata female ses;
run;
proc sort data = samp3;
by female ses;
run;
proc means data = samp3;
by female ses;
run;
* Specifying the number of observations to be sampled in each strata;
proc surveyselect data = "D:\hsb2" out = samp4 method = srs n = (15 20 10 30 25) seed = 9876;
strata female ses;
run;
proc sort data = samp4;
by female ses;
run;
proc means data = samp4;
by female ses;
run;

------------------------------------------------------------------------------