Figure 6.

Model misspecification in a simulation study of stop-screen trials. The graph shows violin plots of maximum likelihood estimates for the parameters w (A), λ (B), and β (C) in a progressive model (i.e., one that assumes $\mathrm{\psi }=0$) fitted to data generated using a mixture model with selected values of $\mathrm{\psi }\ge 0$, assuming that 50,000 women are screened annually at ages 50–54 years with follow-up to age 60 years and that the screening test has 80% sensitivity. Dashed horizontal lines indicate true parameter values. Results were based on 200 simulations per $\mathrm{\psi }$ value.