topic Subsetting Dataset based on number of observations in SAS Programming

Subsetting Dataset based on number of observations

gurpreetkaur — Tue, 03 Jul 2018 11:54:46 GMT

Hi Experts,

I need to create an automation system for creating subsets based on number of observations present in master dataset. For example: if my master dataset has 100 observations, I need to create 4 subsets each having 25 observations. I am unable to think of a logic. If my master dataset has 200 observations then I need to create 4 datasets of 50 observations each. Please help me in writing this logic.

Thank you,

Gurpreet

Re: Subsetting Dataset based on number of observations

RW9 — Tue, 03 Jul 2018 12:05:05 GMT

Get number of observations, then run code, e.g:

proc sql noprint;
  select nobs/4
  into     :cnt
  from    dictionary.tables
  where  libname="<yourlib>"
     and   memname="<yourdataset>";
quit;

data want1 want2 want3 want4;
  set <yourlib>.<yourdataset>;
  if _n_ < (1 * &cnt.) then output want1;
  else if _n_ < (2 *&cnt.) then output want2;
  ...;
run;

Re: Subsetting Dataset based on number of observations

Astounding — Tue, 03 Jul 2018 12:24:56 GMT

If your data source is a SAS data set, you can pull the number of observations and divide them up in the same step:

data subset1 subset2 subset3 subset4;

set have nobs=_nobs_;

if _n_ <= _nobs_ / 4 then output subset1;

else if _n_ <= _nobs_ / 2 then output subset2;

else if _n_ <= _nobs_ * 3 / 4 then output subset3;

else output subset4;

run;

Re: Subsetting Dataset based on number of observations

Ksharp — Tue, 03 Jul 2018 12:57:12 GMT

proc surveyselect data=sashelp.air out=want group=4;
run;

Re: Subsetting Dataset based on number of observations

RW9 — Tue, 03 Jul 2018 13:00:58 GMT

Nice, I always forget the surveyselect function - even though I posted on it some time back myself 🐵

Re: Subsetting Dataset based on number of observations

Ksharp — Tue, 03 Jul 2018 13:07:17 GMT

ME TOO !

remember and forget something everyday .