Fig. 5

Ability of the likelihood ratio test for ΔHEB to detect different levels of change in bias. Simulation results show the fraction of times H₀ was rejected for 10,000 trials with given values of x and n (parameters: α=0.05, \(\mu ^{a}_{1}=\mu ^{b}_{1}=\mu ^{a}_{2} = 100\), r=10). With n≥6 replicates, a 4-fold change is detectable over 95% of the time. However, detecting a 2-fold change more than 95% of the time requires at least 12 replicates