Q: What is the difference bw QRPEM and FOCE?
A: The goal of maximum likelihood methods is to maximize the marginal Probability Density Function (PDF) to obtain estimates for population parameters. Unfortunately, the marginal PDF often cannot be computed analytically due to the complexity of the involved integral. The difference between the QRPEM and FOCE engines is how this problem is solved. Specifically, the QRPEM engine is based on the expectation-maximization (EM) method, which alternates between an expectation step and a maximization step to obtain the population parameter estimates. The FOCE engines involve approximating the required integral and then maximizing the resulting approximate marginal PDF to obtain population parameter estimates.
Is it always the case that there is a small difference between NONMEM an NLME results?
A: If similar engine and settings are used, then NONMEM and NLME usually give similar results. We have posted a large number of NONMEM-NLME model comparisons to our forum that provides evidence that it is typically the case.
Q: Is it possible to create a VPC comparison with Phoenix and NONMEM ?
A: PsN can be used to generate VPC results for NONMEM models. Note that PsN and R are supported in Phoenix framework, and NLME automatically saves VPC results in some worksheets. Hence, it is possible to do such comparison. We will probably demonstrate this in one of our forthcoming webinar sessions.
Q: If "doafter" is not included in the PML code, how would that affect the results?
A: Note that urine is emptied at each observation. Hence, the urine compartment has to be reset to zero right after each observation. This is done in PML through the “doafter” statement. The advantage of this approach is that there is no need to modify the input dataset. Hence, one can use exactly the same input dataset for different models.