I want to sample a population with systematic and cluster sampling methods and use these samples in linear regression modeling. For this, should I do sampling with replacement or without replacement?
The assumption of independence in data is rejected in sampling without replacement! Furthermore, it appears that sampling with replacement is not exact.