Re: Simulation vs. actual data

From: "Nick Holford" n.holford@auckland.ac.nz Subject: Re: [NMusers] Simulation vs. actual data Date: Tue, June 14, 2005 2:48 pm Toufigh, I think you are describing a visual predictive check i.e. simulate from your final model with say 1000 subjects, then calculate the average and 90% empirical prediction interval from the resulting 1000 predictions at each time point. You can plot the average and 90% PI over time and overlay the observed concentrations. About 90% of your observations should lie within the 90% PI. The average vs time can be considered the population behaviour while the 90% PI shows the extent of variability including random effects in parameters and random effects due to residual error. If you think you have linear kinetics you can dose normalize the observed concentrations to some standard dose and do the simulation with this standard dose. This let's you use one plot to display the results and expected profile for all doses you have studied. Nick