Fig. 3
From: Analysis of multicenter clinical trials with very low event rates
Power for scenarios when the number of centers was 5, 50, and 100 and the total number of participants was 200, 500, 1000, and 5000. The ICC was fixed at 0.025, and the participant distribution was skewed across the centers. *The value for the Mantel–Haenszel model was not available because of convergence issues when the number of centers was 100 and the number of total participants was 200