How do I adjust the sample size in PROC SURVEYSELECT?

[esvenson]

Hello, 

I am a new user of PROC SURVEYSELECT in Base SAS 9.4 and I need to set the sample size to be 10% of the total population size.  I don't need to stratify the population.  How would I set the samp size option to execute PROC SURVEYSELECT on only that 10%?  Would it be as simple as the sampsize option below?  Thank you in advance for your help - I really appreciate it!

proc surveyselect data=other_list out=elos1 
sampsize=0.1 
seed=&seed.;
run;

[Reeza]

What happens when you run it? That should tell you if it's right or not. Or are you getting unexpected behaviour?

[esvenson]

Hi Reeza,

I don't know if I should be using sampsize=0.1 or samprate=0.1.  Can you advise which one will yield a 10% sample of the population?

Thanks,

Eric

[Reeza · Accepted solution]

SAMPSIZE generates an error that it needs an integer, so SAMPRATE will be what you need. And 0.1 is 10%. 

[SAS_Rob]

Hi Eric:



You should be using the SAMPRATE= option if you are going to supply it with a sampling rate.

[esvenson]

Is there also a way to have multiple constraints in a PROC SURVEY SELECT like this?  I have a list of factors that all have to be >0 for to meet the criteria for inclusion in the sample population.  I can't find an example of this anywhere.  Thank you again for all your help!

proc surveyselect data=other_list (where = (factor4>0) and (factor5>0)) out=test  samprate=0.1  seed=&seed.;
run;

[Reeza]

The WHERE statement is what I would suggest or pre-filtering the data would also work.

[esvenson]

Reeza, 

I tried this but it was an invalid use of the AND operator.  Is there a way to do a list of multiple constraints in a PROC SURVEYSELECT step so that ALL factors that must be >0 are included?  Thanks!

proc surveyselect data=other_list (where = (factor4>0) and (factor5>0)) out=test  samprate=0.1  seed=&seed.;
run;

[Reeza]

(where = (factor4>0 and factor5>0))

You had brackets too early, closing the WHERE condition, I think. Try the above.

Unfortunately you're stuck listing the factors but you can check if the max is 0 and that should be enough?

Untested.

(where = (not max(factor1, factor2, factor3, ... factor9) = 0));