Quote:
Originally Posted by elphenor
You use it when you intend to use the data to figure how individual groups voted
let's say you conduct a poll of 1000 Americans, that is a large enough sample size to produce a small margin of error, but mathematically you may only end up with 100 Hispanics for instance not a large enough sample size to get a reasonable margin of error for that specific group
So if they plan to use the data that way they will oversample smaller groups like Hispanics so they get a reasonable amount and then weight them back down when they use the data for other purposes
Sampling | Pew Research Center
|
I was under the impression that oversampling was when you disproportionately sample from one group without it being necessary for the data. What that link describes just sounds like stratified sampling to me.