Table 1.

Summary of different designs for the lavender scent example.

Design	Sample size	Power
Individual randomized trial	$k=1$ cluster, $n=340$ subjects per cluster $k=1$ cluster, $n=1228$ subjects per cluster	0.31 0.80
Cluster randomized trial parallel groups	$k=20$ clusters, $m=157$ subjects per cluster	0.80
Cluster randomized trial Crossover, cross-sectional	$k=20$ clusters, $m=36$ subjects per cluster-period	0.80
Cluster randomized trial Crossover, cohort	$k=20$ clusters, $m=26$ subjects per cluster-period	0.81
Cluster randomized trial Crossover, cohort 25% attrition of subjects	No repair: $k=20$ clusters, $m=26$ subjects in cluster-period 1 Increase m: $k=20$ clusters, $m=30$ subjects in cluster-period 1 Replacement: $k=20$ clusters, $m=26$ subjects per cluster-period	0.75 0.80 0.79