Skip to content

Draft: Treatment models

Pierre-Emmanuel Poulet requested to merge treatment_modelling into dev

What does the code in the MR do ?

The MR provides new model classes with support tools to model Leaspy with disease-modifying treatments. The model is able to create a piece-wise geodesic trajectory for an individual given certain breakpoints which corresponds to the change of directions in the trajectory. Breakpoints can be thought of as changes in treatment strategy (new treatment, dose adjustment...) but they do not correspond necessarily to the visits : there can be several visits between two successive breakpoints.

Currently the implementation supports linear and logistic models.

Where should the reviewer start ?

Review should not start now by any means.

The reviewer should start by the univariate_treatment_model file, then move on to the models/utils/treatment_functions file where the current implementation of the function generating the changes in trajectories is proposed.

How can the code be tested ?

At this moment tests have not been added, but one can open a notebook, generate data with a 'univariate_treatment_linear' Leaspy model for instance. The code to fit a model is no different than a normal Leaspy.

When is the MR due for? (review deadline)

End of december probably.

What issues are linked to the MR ?

None

Edited by Pierre-Emmanuel Poulet

Merge request reports