Simultaneously estimating IIV and IOV estimate for first cycle in pharmacokinetic NLME model using Torsten

[marianklose]

@Bob_Carpenter

@jd_c put their finger on the problem. You can see the non-identifiability here. The likelihood doesn’t identify ETA_CL or KAPP_CL, only the prior does. We discuss exactly this case in the problematic posteriors chapter of the User’s Guide.

Thanks for pointing this out! I will have a look again at the User’s Guide. I guess it makes sense that only the prior allows for some distinction of ETA_CL and KAPP_CL, but the likelihood only informs CL itself, not necessarily each of the two parts.

@Bob_Carpenter

Is that conditional a typo? The two branches are identical. If that isn’t a typo, there’s no need to define these auxiliary parameters like KAPPA_CL, which is always identical to KAPPA_CL_C1.

Well spotted. Not really a typo, the model is rather at a debugging state. I have decided to first get it running with CYCLE = 1 data only and then adapt it for multiple cycles. NONMEM typically complains when you have conditionals without a general else block in the end, so I guess it is just a habit. But I got already some good input in Dynamically specifying which parameters to estimate in stan how to do it without a conditional.

@Bob_Carpenter

That explains the all-caps Fortran-style variable naming convention!

Yep, a good way to spot a NONMEM / Fortran user ;-)