I have a data challenge where I have a specific individual, and several data have been collected relating to several of its characteristics. It is a part of large data frame with hundreds of thousands of rows, and it is my job to be able to select about 20-30 peers from within this gigantic table for the specific individual.

My biggest challenge is how to narrow the list down. There are a couple obvious categories to filter based on, but those still return thousands of peers.

Another issue is that some rows have some columns populated, some are missing. So if I went to filter based on a sparsely-populated variable, I would be self-selecting only for peers that have that column filled in, which I think would be considered selection bias.

Another idea that I had was to just pick a random sample of 50 (or some number around that) assuming it would be representative of the general population but I’n not sure if that is statistically sound.

Another thing of note is that all variables are categorical.

Anyone have other suggestions?