In PBPK context when the model is defined by set of ODEs the time to estimate joint posterior takes quite a while (in my case ~ 3 days). The idea is to estimate joint posterior Ppop using the data for N subjects to determine population estimate (9x14 data points, hierarchical model Mpop) and then use this estimate as a prior but using data for only N+1 subject (2 data points, the hierarchical model condensed to one subject non-hierarchical model Mind) to estimate Pind. Is this possible using stan?
What I typically do is to use all the data (9x14+2 data points) and hierarchical model Mpop. However, this is somewhat inefficient since in order to evaluate which time points for the concentration profile of N+1 subject are most informative I have to run hierarchical model Mpop multiple times. Ideally I would like to estimate Ppop once by running Mpop and data for N individuals and then run Mind multiple times using two data points for N+1 individual.
Thanks for any suggestions.