Skip to main content

Table 2 Simulated type I error rates of the testing approach in specific scenarios

From: Hypothesis testing in Bayesian network meta-analysis

n ik piP, piS Δ contrast-based arm-based
1000 0.2 log(0.9) 0.024 0.024
   log(1) 0.024 0.024
   log(1.1) 0.024 0.025
200 0.1 log(0.9) 0.023 0.025
   log(0.95) 0.022 0.025
   log(1) 0.022 0.026
   log(1.05) 0.021 0.025
   log(1.1) 0.023 0.026
400 0.05 log(0.5) 0.032 0.024
   log(2) 0.017 0.028
  1. nik denotes the number of treatment arms, piP and piS the success rates in arm P and S, respectively, in trial i (i=1,…,16), and Δ is the non-inferiority or superiority margin, respectively. We used 50,000 simulated data sets to estimate the type I error rate. The nominal level of α was 0.025