Designing Minimum-Cost Multi-Stage Sample Designs

• Activity Number: 514
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Matthias Ganninger*+
• Companies: GESIS - Leibniz-Institute for the Social Sciences
• Keywords: design effects ; multi-stage design ; costs ; minimum

Abstract

In cross-national sample surveys like the ESS, a huge variety of sample designs is often applied in participating countries. In order to achieve estimates of comparable precision, the samples drawn according to these different sampling schemes must be of equivalent effective sample sizes, n_eff = n/deff, where n is the net sample size and deff is the design effect. As deff, among another parameter, depends on the average cluster size b, increasing the number of sampled clusters, ceteris paribus, decreases the design effect and hence increases n_eff. The presentation will show that, at a given linear cost structure (costs per interview and costs per sampled cluster), there exists an optimal number of clusters to sample so that a pre-defined effective sample size is exactly achieved - at minimum total costs.