Version 2.0.1

• New: Pipeline.build()

  • This function just essentially calls get_estimator(), but may be more intuitive to some to have it be accessible via the Pipeline classes.

• New: Evaluator.subset_by()

  • Post stratifying / splitting results by different subsets of subjects, i.e., by sex, can be a useful tool in performing post-hoc diagnostic tests on models. This new method and outputted classes tries to make this process easier.

• Changed: train_indices and val_indices in BPtEvaluator to train_subjects and val_subjects

  • This change was made so that parameter names can be more consistent with the Subjects style input and other naming schemes.

• Minor Improvement: Added warning when invalid extra_params passed

  • Now when extra params are passed to functions like evaluate() if the keys don’t line up with any valid parameters, a warning will be issued.

• Minor Improvement: Can now pass single model to Pipeline

  • Before you had to pass steps as a list, but now Pipeline will accept a single model-like step as input.

• Updated: Predicting Sex Example

  • Added some more examples of using different more advanced features.

• Fixed Tests so Windows Compatible

  • Had hard code file paths that were not compatible with windows before