English title: Designing minimum-cost multi-stage samples for cross-national surveys
Author(s): Matthias Ganninger -
Type: Conference paper/poster
The design effect is a measure for the inflation of variance of an appropriate estimator for a population parameter under a complex sample design compared to the variance of an appropriate estimator for the same parameter under simple random sampling with replacement. In the case of two-stage sampling with unequal inclusion probabilities, the design effect can be decomposed to two components, deff_p and deff_c. Under the assumption of a linear cost model, this model-based approach to design effects can be used to predict an optimal number of clusters to sample at minimum total cost and a condition on the precision of an estimator.
Conference name: NTTS
Start date: Feb 22, 2011