The Design Effect (DE) is calculated for clustered samples because each element in the cluster is not completely independent of other elements in the cluster.
Synonym: Inflation Factor
Parameters:
(1) average size of the cluster (number of patients, elements, etc)
(2) intra-class correlation (ICC), a decimal fraction from -1 to +1 (+1 for maximal homogeneity of cluster elements and -1 indicates maximal heterogeneity)
design effect =
= 1 + (((number of things in the cluster) - 1) * (intra-class correlation))
sample size if clustered =
= (design effect) * (sample size if randomized)
The sample variance will be (the design effect) times larger with the clustered data than it would be if the study was done with randomly selected sample of the same size.
To read more or access our algorithms and calculators, please log in or register.
