davaug / erlotinibgefitinib Goto Github PK

Allows distributing the parameters of a structural model with population distirbutions.

Analogously to pkpd.optimise

Binder's container need sundials in order for the notebooks to be fully functional.

The inference of the structural + error model can be straightforwardly performed using pints.ProblemLogLikelihoods. However, forward simulation of the error appears to be rather cumbersome with the Likelihoods.

A straightforward solution is to implement Error classes for forward simulation by wrapping scikit statistical models. Those models should be able to produce random samples as well as compute percentiles.

With this issue also the corresponding pkpd.plots.plot_error function should be refactored.

Wrap what is in the error model notebook into a function.

Create pints model for the tumour growth in absence of treatment

Implemement a pints.LogPDF that allows to share the noise parameters between problems.