InsufficientGibbs 1.0.4

Package for performing Bayesian inference with insufficient and robust statistics

Insufficient-Gibbs-Sampling

InsufficientGibbs is a package that enables users to sample from the posterior parameters when only certain robust statistics of the data are accessible. The paper that comprehensively describes the theory behind these methods can be found on arXiv (https://arxiv.org/abs/2307.14973). Additionally, the code responsible for generating all the paper figures can be located in the figures folder.

Robust Gibbs package

We propose here three main functions named Gibbs_med_MAD, Gibbs_med_IQR and Gibbs_Quantile to cover the case when we observe the pairs (median, MAD) or (median, IQR) or a sequence of quantiles.

Install

Install latest release via pip

pip install InsufficientGibbs

For latest development version clone the repository and install via pip

git clone https://github.com/antoineluciano/Insufficient-Gibbs-Sampling
cd InsufficientGibbs
pip install .

Available models/likelihood

• Normal distribution (distribution="normal")
• Cauchy distribution (distribution="cauchy")
• Weibull distribution (distribution="weibull")
• Translated distribution (distribution="translated_weibull")

Available location priors

• Normal (par_loc="normal")
• Cauchy (par_loc="cauchy")
• Gamma (par_loc="gamma")

Available scale priors

• Gamma (par_loc="gamma")
• Jeffreys (par_loc="jeffreys")

Available shape priors

• Gamma (par_loc="gamma")