1) We should investigate how nlmixr handles mixture models.
2) Probably, mixture models are simply defined as a list of probable models, with an a priori chance of belonging to subpopulation 1 or subpopulation 2.
3) The mixture probability a posteriori should be estimated (based on log-likelihood + a priori chance) and carried forward into simulations.
Will largely be an API definition challenge, to make it very clear to the pharmacometrician what we are doing.
rfaelens
commented
5 years ago
Allow the definition of mixture models.
1) We should investigate how nlmixr handles mixture models. 2) Probably, mixture models are simply defined as a list of probable models, with an a priori chance of belonging to subpopulation 1 or subpopulation 2. 3) The mixture probability a posteriori should be estimated (based on log-likelihood + a priori chance) and carried forward into simulations.
Will largely be an API definition challenge, to make it very clear to the pharmacometrician what we are doing.