topic Re: Splitting data in SAS Procedures

Splitting data

Greeshma — Sat, 28 Apr 2018 05:29:46 GMT

How to split the data into three randoms sets: 60% in a building set, %10 in a test set and 30% in a validation set.

Re: Splitting data

PGStats — Sat, 28 Apr 2018 05:50:06 GMT

You can use the GROUPS= option in proc surveyselect. You must provide the exact set sizes of calculate them like this:

proc sql;
select 
    round(0.6*count(*)) as n1,
    round(0.1*count(*)) as n2,
    count(*) - calculated n1 - calculated n2 as n3
into :n1, :n2, :n3
from sashelp.class;
quit;

proc surveyselect data=sashelp.class groups=(&n1 &n2 &n3) out=samples;
run;

Re: Splitting data

Ksharp — Sat, 28 Apr 2018 10:51:41 GMT

data train validate test;
 set sashelp.heart;
 call streaminit(123456789);
 x=rand('table',0.6,0.3,0.1);
 if x=1 then output train;
  else if x=2 then output validate;
   else output test;
drop x;
run;

Re: Splitting data

Greeshma — Sat, 28 Apr 2018 17:18:25 GMT

Thank you, that helped.

Re: Splitting data

Greeshma — Sat, 28 Apr 2018 17:18:55 GMT

Thank you