|
Abstract:
|
Selection of the primary estimand and corresponding analysis methods in the presence of missing response information is a complex problem. This presentation uses neuroscience clinical trial examples to describe how simulations can be a powerful tool to address this complex problem at the trial design stage. The performed simulations incorporate various assumptions for the trial. Simulation metrics such as the estimated power and Type I error rate, mean and standard error for the treatment difference versus control, etc. allow a quantitative comparison of the performance of different methods. The talk will show how the simulation results play a critical role towards an informed selection of the primary estimand, primary analysis method and additional sensitivity analyses.
|
