Fig. 3From: Hypothesis testing in Bayesian network meta-analysisPower values. Power for the arm-based and contrast-based approach for a varying sample size Ni,k (based on 10,000 runs). The success rate was kept fixed at pik=0.1 while the number of observations was varied (i=1,…,16;k=P;S;E). An OR of 1.15 was used for power simulation while Δ=0 was usedBack to article page