Figure 4
From: Assessing potential sources of clustering in individually randomised trials
Simulation results (50 therapists, 1000 patients). Continuous outcomes were generated based on a treatment effect of 0 and therapist effects and a random error term, both of which followed a normal distribution. Patients were assigned to therapist post-randomisation. An equal number of patients were assigned to each therapist, and we used 5000 replications for each scenario.