Samling procedures

[Biniie]

Hi

I have a big dataset from which I want a specific subsample. 

The dataset contains workshift data from employees from approx 100 different companies, which have different number of employees. The dataset can include selveral observations/datalines for each person.

I want a subsample that consist of all the observations for 10% of the 100 companies (and not 10% of the number of observations/datalines), where the 10% of the companies is randomly chosen. (I'm aware that this means, that the subsample could be of different size depending on which companies, that are chosen). 

Anyone who has a suggest how to do this?

[PaigeMiller]

PROC SURVEYSELECT with the STRATA statement can do this.

[FreelanceReinh]

Hi @Biniie,

The CLUSTER statement (alias: SAMPLINGUNIT) is ideal for your purpose.

Example:

proc surveyselect data=workshiftdata
method=srs samprate=10
seed=2718 out=subsample;
cluster company;
run;

(Just replace workshiftdata, subsample and company by your input dataset, output dataset and variable names, respectively.)