aloctavodia / kulprit

Kullback-Leibler projections for Bayesian model selection in Generalised Linear Models

kulprit

(Pronounced: kuːl.prɪt)

Getting Started | Documentation | Contributing

---

Kullback-Leibler projections for Bayesian model selection in Python.

Example workflow

import pandas as pd
import numpy as np

import bambi as bmb
import kulprit as kpt

import arviz as az
import matplotlib.pyplot as plt

# define the data
x = np.random.normal(0, 0.25, 121)
y = np.random.normal(0, 1, 121)
z = np.random.normal(2 * x + 3, 0.1)
data = pd.DataFrame({"x": x, "y": y, "z": z})

# fit the reference model
model = bmb.Model("z ~ x + y", data, family="gaussian")
idata = model.fit()

# build reference model object and perform the search procedure
ref_model = kpt.ReferenceModel(model, idata)
ref_model.search()

# compare the projected posterior densities of the submodels
ax = ref_model.plot_densities();

# compare submodels found in the search by LOO-CV ELPD
cmp, ax = ref_model.loo_compare(plot=True);
cmp

# project the reference model onto a chosen submodel size
submodel = ref_model.project(1)

# visualise projected parameters
ax = az.plot_posterior(submodel.idata, var_names=submodel.model.term_names);

Installation

Currently, this package is only available for download directly from GitHub with the command

$ pip install git+https://github.com/yannmclatchie/kulprit.git

Development

Read our development guide in CONTRIBUTING.md.

---

This project was generated using the biscuit cookiecutter.