topic Re: Sampling without replacement, create two data sets in SAS Data Management

Sampling without replacement, create two data sets

desireatem — Wed, 07 Mar 2018 17:08:17 GMT

I have this data set A with N=100. I want to create a data set B with N=75 through simple random sampling without replacement and the remaining 25 should be data set C.

Help!

Re: Sampling without replacement, create two data sets

ballardw — Wed, 07 Mar 2018 17:24:27 GMT

proc surveyselect data=have out=sampled 
     sampsize=75 outall;
run;

will create a data set sampled with all of the records from the have set and a variable Selected to indicate whether it was selected for the 75 or not. If you want a percentage of records then use SAMPRATE instead of SAMPSIZE.

If you really need to split the data then a separate data step

data samp75 samp25;
  set sampled;
  if selected then output samp75;
  else output samp25;
run;

Re: Sampling without replacement, create two data sets

Patrick — Wed, 07 Mar 2018 23:32:53 GMT

Beginning with SAS 9.4 Proc Surveyselect allows you to directly create both data sets.

http://support.sas.com/kb/36/383.html

In case you haven't a SAS license which gives you access to Proc Surveyselect: There is also a data step approach possible. Just amend the code in Method 3 under this link http://support.sas.com/kb/24/722.html to write non selected rows to another table.