i've got questions about clustered-stratified random sampling.
let's say I want to do a research in narcotics.
the population that I will be working with is a combination of academics, students, and government employees (law enforcers, DEA, etc). I consider the population's unknown because I couldn't get the exact number of the population.
from that population, I'm thinking of clustering the population into two groups, those from academics and the practitioners (the government employees). From that clusters, I developed two levels,
the academic cluster contains lecturer group and students group.
the practitioners group contains high-level group and low-level group.
Have I already clustered and stratified my sample in right manner?
How do I decide the exact sample for the cluster without knowing specific population of each group and without knowing total population?
and how do I decide the exact sample for each level of the cluster?
thank you very much, I appreciate your time and suggestions 🙂
Best Answer
What you described is stratification: you know before you sample that a given unit is a professor, or a student, or a law enforcement officer. If you know something about an observation unit beforehand, that's typically is (or can be) a stratification variable. Now, the clusters would be units you would sample together for logistics reasons: you don't have a full list of students in your country, but you have a full list of universities, and you can sample may be 20 of these, and try to reach students or professors in these universities somehow (that's difficult, but I will leave these difficulties to you). Now, university is then a cluster. Within that cluster, you stratify your potential respondents into professors and students, and take samples of these independently. So you have multiple complex sample features:
There will be tricks along the way -- you would want to sample proportional to size when taking clusters, for the reasons you can find in any good sampling book.