chapter  3
Choosing Whether to Randomise by Cluster
Pages 18

We have seen in Chapter 2 that correlations between observations on individuals within the same cluster lead to estimates that are less precise than those based on simple random samples. In a CRT, we collect data on a sample of clusters in each treatment arm, and our estimates of treatment effect are based on comparison of the results from these cluster samples. It follows from this that, in general, a CRT will provide less precise estimates than an individually randomised trial with the same total number of individuals.