This Python package provides implementations for the Horseshoe and Horseshoe+ prior calculations found in "The Horseshoe+ Estimator of Ultra-Sparse Signals", " Default Bayesian analysis with global-local shrinkage priors" and "Prediction risk for global-local shrinkage regression".
Specifically,
- Generic numeric and symbolic functions--via
mpmathand SymPy, respectively--for bivariate confluent hypergeometric functions. - Expectation and moment calculations for the hypergeometric inverted-beta distribution.
- SURE values.
Install directly from the repository with the following:
$ pip install git+https://github.com/brandonwillard/hsplus-python
After cloning the Git repository, use the -e option in the project directory:
$ pip install -e ./