RE: Simulation vs. actual data

From: "Perez Ruixo, Juan Jose [PRDBE]" JPEREZRU@PRDBE.jnj.com Subject: RE: [NMusers] Simulation vs. actual data Date: Thu, July 14, 2005 7:31 pm Ken, Thanks for your well detailed explanation. I understand to get N sets of population parameters (thetas, Omegas, and Sigmas) from a non-parametric BS in order to simulate N clinical studies is computationally burdensome, specially for large N (let say 2500). I think the key question is how large N should be to account properly for the uncertainty. The purist way says that we should run SUBPROBLEM=1 for each set of population parameters. However, if a smaller N is able to account properly for the uncertainty then the computational burdensome is less. Suppose, 50 non-parametric BS replicates are good enough to account for parameter uncertainty, then would you consider to use 50 sets of population parameters and run SUBPROBLEM=50 in order to get 2500 trial replicates? Regards, Juanjo.