v2.8.0

Merged PR :

• #69
• #70

Typing :

• Typing non parametric model with Generic : it allows to specify available estimations in concrete classes.
• Refactoring utils and remove relife.type.
• Typing Numpy i/o with optype shaped arrays : probability functions are typed as if they can expect ArrayND and carry out potential broadcasting correctly. For now, it is not true at runtime but a future PR will solve it.

API changes :

• Refactoring plot method : model.plot("<fname>", time, *args, **kwargs) instead of model.plot.<fname>(*args, **kwargs).
• Modifying module names to plural when it makes sense : e.g. lifetime_model -> lifetime_models, doc -> docs, stochastic_process -> stochastic_processes, etc.
• Rename module data to datasets (more explicit) and delegate NHPPData to non_homogeneous_poisson_process module.
• @params.setter is replaced by set_params because @property-params does not have the same definition. Same for @property-params, it is replaced by get_params.
• Removing nb_params (useless).
• Removing shape/rate attributes (useless) in distributions.
• same for ar, cp, cf, etc. in policies.
• Rename utils functions : to_column_2d and flatten_if_at_least_2d are more explicit.
• Remove FrozenParametricModel. It is useless. Now, FrozenParametricLifetimeModel derived directly from ParametricLifetimeModel[()]. It simplifies typing in policies and stochastic processes as LifetimeDistribution is a ParametricLifetimeModel[()], thus is accepted by default as policies input.
• Remove get_model_nb_assets(model) from utils (useless and avoid circular imports).
• Rename reshape_1d_arg to to_column_2d_if_1d (more explicit).
• Removing nested tests in stochastic_process : all tests are done in one tests folder.
• Use plural module names if it makes sense.

Important choices :

• For typing, optype.numpy.ArrayND is prefered to numpy.typing.NDArray.

Docstrings :

• If docstrings are too long, when it's needed, add # noqa: E501.

Future works :

• Solve Numpy broadcasting in every models' functions.
• Typing pytests
• Refactoring tests

v2.7.1

• Documentation update
• Removing doc group

v2.7.0

• Simplify rvs integration with conditional model. PR #67
• Bugfix : supporting 1d array of covar values in fit for parametric regression models if one covariate is expected by the model

v2.6.3

BUGFIX : b576acf

v2.6.2

BUGFIX : 173e3b0

v2.6.1

CI changes : 40656b7

v2.6.0

Adding Cox regression model : 158fba5
More documentation will be added later.

v2.5.2

bugfix 4c944e9

v2.5.1

bugfix e35d02f

v2.5.0

Adding KijimaProcess (first code) : PR #64