Take a random sample of unique values of a variable

[Arthur Carvalho Brito]

Hello,

How can I take a sample of unique values of a variable, and then filter my dataset according to those drawn values?

That is, assume my variable product contains multiple product identifiers (strings). More than one row of the data is attached to a given product, and different products appear in different frequencies in the data.

(therefore, simply taking a sample of the data itself won't suffice).

How can I take a sample of size N of the product identifiers and then filter the data to keep observations linked to those identifiers that were drawn?

I tried using something with

levelsof product, local(unique_values)

but didn't know how to use the sample command with a local

Thank you

[Joseph Coveney]

Maybe something along the following lines.

frame copy default Selection
frame Selection {
    contract product
    quietly sample N, count
}
quietly frlink m:1 product, frame(Selection)
quietly keep if !mi(Selection)

You'll need to (i) set the seed in order to help assure reproducibility and (ii) change N above to the sample size you want.

[Nick Cox]

Please see Section 2 of https://journals.sagepub.com/doi/pdf...867X0800800408 for a recommendation to use the term distinct, not unique, for the meaning intended here.