Skip to main content


Fig. 2 | BMC Medical Research Methodology

Fig. 2

From: The impact of varying cluster size in cross-sectional stepped-wedge cluster randomised trials

Fig. 2

Variation in power for a cross-sectional SW-CRT with six clusters and six sequences when the cluster sizes are a) known and b) not known. a The clusters sizes are known pre-trial and are unequal in size – the cluster-period sizes are: 15, 25, 35, 45, 80 and 100. The power has been calculated for each possible randomisation order of the six clusters using the methods derived by Hussey and Hughes. b The clusters sizes are unequal but not known, but it is expected that the average cluster size per cluster-period = 50, and the coefficient of variation = 0.66. Potential cluster-period sizes have been simulated and used to estimate the power using the methods derived by Hussey and Hughes. This has been repeated 4000 times, with each simulation containing new cluster-period sizes. The SW-CRT contains six general practices randomised over six sequences, with one general practice crossing from the control to intervention condition every month. The SW-CRT is designed, such that the difference to detect = 0.2 (SD: 0.8); type 1 error = 5%

Back to article page