subsetting random observations

[deleted_user]

Hi,

I have a data set in which i want to subset a sample with a condition. In the subset I want to pull out 1.5 % of random observations of a variable and take the average of it.

I could do the first part of this task to pull the subset with a condition (using WHERE clause).

I'm trying to figure out how I can pull 1.5 % of random observations of the variable.

After that, I think I can use Avg function to get a mean of it.

Appreciate you help!

[Olivier]

Hi.
Check out the Surveyselect procedure if you have licence for SAS/STAT.
If not, one usual way is to code :
1) a data step that creates a random variable (using the RANUNI or RAND functions)
2) sort the new dataset by this random variable
3) keep only the x first observations of the sorted dataset.
This last way of doing is just like shuffling cards.

Olivier

[LinusH]

You can also do a subsetting if in the first datastep (on the new "ranuni"-variable). Then there is no need to sort and read the data a third time.

Regards,
Linus

[Patrick]

The code below (second data step) is almost what SAS provides in the training for the SAS certification.
HTH
Patrick

data work.rsubset;
set (where=( run;

data work.rsubset(drop=obsleft sampsize);
/* sampsize=10; */
sampsize=ceil(totobs*0.015);
obsleft=totobs;
do while(sampsize>0);
pickit+1;
if ranuni(0) set sasuser.revenue point=pickit
nobs=totobs;
output;
sampsize=sampsize-1;
end;
obsleft=obsleft-1;
end;
stop;
run;

Message was edited by: Patrick Message was edited by: Patrick

[mkeintz]

San:

Are you asking for 1.5% randomly selected from the entire data set?  Or do you want 1.5% of the subset that meets a given condition (you said "subset a random sample with a condition").  If it's the latter, then your question hasn't been answered yet.

[ArtC]

We should also ask whether the subset is selected with or without replacement.  For small subsets of a large data set it probably does not matter a lot, but....