Power to detect different effect sizes (ES) by coefficient of variation (CV) in enrolment given randomisation of 20 sites
| Effect size (difference in proportions) | ||||||
| Coefficient of variation in enrolment | 0.05 | 0.06 | 0.07 | 0.08 | 0.09 | |
| 0 | 87% | 94% | 97% | 99% | 99% | |
| 0.2 | 86% | 93% | 97% | 99% | 99% | |
| 0.5 | 80% | 89% | 94% | 97% | 98% | |
| 0.8 | 70% | 80% | 87% | 92% | 95% | |
| 1.0 | 62% | 72% | 80% | 87% | 91% | |
Even with large variability in participant enrolment (CV=1), we will have over 90% power to detect a difference of 0.09. We will have good power (>80%) to detect effect sizes as low as 0.05 provided the variability in site enrolment is not great (<0.50).
Grey represents power greater over 90%, light grey represents power between 80–90%, and medium grey represents power <80%.