I'm trying to understand how sampling design affects analyses and I'm a little confused about how to adjust for clustering. From what I've read, when you have a clustered sample you are supposed to calculate (or estimate from previous research) a design effect for use in both the power calculation and the standard error for any hypothesis tests.
In cases where researchers are only interested in the micro-units (i.e. they don't want to make inferences about the clusters) and they thus choose not to use a multilevel model, I've noticed that most seem to just include dummy-coded regressors in a multiple regression model to control for the effect of cluster.
Is this sufficient, or is it inferior to adjusting the standard error in the analysis by the design effect?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.