How small is too small for a subsample?


I’m using the CPS basic monthly and the ASEC. My analyses involve several subpopulations… for example, respondent is a woman and a racial minority and a business owner. How small is too small for the data to be reliable when talking about the count of observations that go into an analysis? Thank you!



There is really no bright line criteria for a sample size that is “too small.” With that said, you can calculate some measure of the margin of error associated with your population counts. Then, based on your ultimate objectives for the analysis, you can make a decision about how large of a margin of error you are comfortable with. If you are using a statistical software, then most standard packages will automatically calculate standard errors or standard deviations with population estimates.



Awesome, Jeff. Thank you. That was my understanding as well, but I paused and figured it was best to ask the community of experts!