Bootstrap sampling from a long formatted file

[BlueNose]

Hello all,

I have a long formatted file, in which every subject has more than one row. For each row, I have several variables, two of them are "ID" and "Visit". Each subject has 5 visits. I have 50 subjects.

I want to make a random sampling with replacement, in order to get a larger file, with 100 subjects.

The problem is, when I use PROC SURVEYSELECT (which is straightforward), SAS chooses rows, not subjects. I want SAS to sample a subject, and to keep all his rows, rather than randomly selecting rows.

Can it be done ? I am pretty much lost here.

Thank you.

[Kurt_Bremser]

Have you already tried to use proc transpose to reformat the file into rows?

[BlueNose]

No. I have too many variables per visit (10 variables * 5 visits will give 50 variables wide).

[BlueNose]

I have an idea. Can I save the number of times each subject is selected, and then build different datsets for subjects with number of hits of up to 1, up to 2, etc... and then append these datasets ?

[data_null__ · Accepted solution]

• SAMPLINGUNIT | CLUSTER variables < / options > ;