Does variance tend to be higher in cluster sampling compared to SRS? Why?
The cluster sampling design usually leads to a larger sampling variance
than that from SRS. This is because the elements within naturally formed
clusters are often similar, which then yields a positive covariance between
elements within the cluster. The homogeneity within clusters is measured
by the intraclass correlation coefficient (ICC)—the correlation between all
possible pairs of elements within clusters. If clusters were randomly formed
(i.e., if each cluster were a random sample of elements), the ICC would be
zero. In many natural clusters, the ICC is positive and, hence, the sampling
variance will be larger than that for the SRS design
Get Answers For Free
Most questions answered within 1 hours.