Diagrams showing how different assumptions about bias proportions change the confidence interval for the effect of Standard to Placebo in the setting of the non-inferiority trial. In each of the three panels A, B, and C, the lined region (on the right side of each panel) represents the 95% confidence interval for βPS under different assumptions about λ, the amount by which β̂PS,H is biased for βPS. The value of τ2 in panel C maintains the width of the confidence interval from panel A, reflecting uncertainty about the true value of λ.