Bambi

Maintenance and fixes

- Change default priors for the 't' family (403)

Documentation

- Add installation instructions with conda (406)
- Corrected a typo: pary_id -> party_id (402)
- Add donation information (409)

Maintenance and fixes

- Documentation for all versions is built from scratch when there's a release. This ensures older versions link to the current stable release. (396)
- Add new axis to prior predictive samples to represent 1 chain in the InferenceData object we return (397)
- Move Family, Likelihood and Link to the families submodule and improved some docstrings (399)

Documentation

- Add example with hierarchical binomial model (398)

New features

- Add alternative default priors (360)
- Add StudentT family (367)
- Add Beta family (368)
- Implement both in-sample and out-of-sample model predictions (372)
- Add function to load datasets (375)
- Add option to specify potentials (379)
- Add Binomial family (386)

Maintenance and fixes

- Automatic switch initialization method from "jitter-adapt_diag" to "adapt_diag" when sampling fails (383)
- Predictors are internally centered when there is an intercept. This generally results in improved sampling efficiency (385)
- Improve documentation and error message in `Model.graph()` (390)

Documentation

- Add Negative Binomial family example notebook (346)
- Fixed typos and improved many notebooks (374, 377, 382)

New features

- It is possible to specify priors for parameters in the response distribution (335)
- Add probit and cloglog link functions (340)

Maintenance and fixes

- Informative message when default priors fail because of perfect separation. Model can be fit with custom priors (330)
- Breaking changes to the API. All the information related to the model goes in `Model()` instantiation now (333)
- Fix gamma family (337)
- Non-default links are properly passed to statsmodels (337)
- Fix Wald family (340)
- Fix Negative binomial family (340)
- Add informative message when link function is not available for a given family (340)
- Update formulae version to 0.0.10 (348)

Documentation

- Notebooks are updated to the new API (336)
- Add badges, update introduction and minor style changes in webpage (344)
- Add example using Gamma and Wald families (345)
- Webpage theme has been updated to PyData theme (347)
- Add model evaluation to logistic regression example (350)

New features

- Add option to save a figure from model.graph() by passing the name of a file. Figure format and resolution can also be set (317)
- Objects of class Prior, Family and Model have nicer print methods (326)

Maintenance and fixes

- Add negative binomial family to config file, which was missing (324)
- Add test to check model compilation with families available (327)
- Update formulae to version 0.0.9 (329)

Documentation

- Fix gamma docstring (328)

New features

- Use formulae to parse model formulas (299)
- Add model representation (300)

Maintenance and fixes

- Remove deprecation warning related to pm.sample returning idata (295)

Documentation

- Add citation to Bambi preprint (290)
- Remove reference to pystan (292)