Fig. 3.
Coverage of estimators across different treatment effect and non-enrolment scenarios for simulation study 2a. Error bars are 95% confidence intervals based on Monte Carlo standard errors. Treatment effect scenario 1: Constant treatment effect. 2: Treatment effect varies across episode. 3: Treatment effect varies across patients with different values of Mi. 4: Treatment effect carries forward. 5: Treatment becomes less effective on re-use. 6: Treatment effect varies across episodes, across patients with different values of Mi, carries forward, and becomes less effective on re-use. Non-enrolment scenario 1: non-enrolment depends on previous treatment allocation. 2: Non-enrolment depends on previous treatment allocation and previous outcome. 3: Non-enrolment depends on previous treatment allocation and baseline prognosis at episode two. 4: Non-enrolment is differential between treatment groups based on previous outcome. 5: Non-enrolment is differential between treatment groups based on baseline prognosis at episode two