Estimating Marginal Likelihoods Using Bridgesampling

This repository contains the code and data for a paper proposing diagnostic methods for bridgesampling. The experiments are divided into two main parts: analysis using posterior distributions from the posteriorDB project and a toy example exploring the effects of increasing the number of covariates in a generalized linear model (GLM).

Repository Structure

• posteriordb/: Contains the main experiments that apply bridgesampling to various posterior distributions curated from the posteriorDB. This folder includes both R scripts and Stan models that are essential for replicating the findings and further exploration.
• toy_example/: Includes experiments designed to study the impact of an increasing number of covariates in a GLM on the efficiency and accuracy of the bridgesampling method. This section is helpful for understanding scalability and performance in simpler, controlled scenarios.

Contents

• R files: Scripts for setting up the statistical models, executing the bridgesampling algorithm, and processing the results.
• Stan files: Stan models used to define the Bayesian models whose posteriors are being explored.
• Data files: Data files used in the experiments, stored in csv format.

Getting Started

Prerequisites

Ensure you have the following software and libraries installed:

• R (version 4.0 or higher)
• RStan (version 2.21 or higher)
• bridgesampling package in R
• CmdstanR (bridgesampling version)

Installation

Clone the repository to your local machine using:

git clone https://github.com/GiorgioMB/bridgesampling_paper_code.git
cd bridgesampling_paper_code

Usage

Navigate to either of the experiment directories and run the R scripts provided. For example:

cd posteriordb
Rscript low_dim_gauss_mix.R

This will execute the analysis using the predefined model and data from posteriorDB.

Contacts

For any queries related to the repository, please contact:

ext-giorgio.micaletto@aalto.fi