Description

The recent availability of huge, many-dimensional data sets, like those arising from genome-wide association studies (GWAS), provides many opportunities for strengthening causal inference. One popular approach is to utilize these many-dimensional measurements as instrumental variables (instruments) for improving the causal effect estimate between other pairs of variables. Unfortunately, searching for proper instruments in a many-dimensional set of candidates is a daunting task due to the intractable model space and the fact that we cannot directly test which of these candidates are valid. We propose a general and efficient causal inference algorithm (MASSIVE) consisting of Model Assessment and Stochastic Search for Instrumental Variable Estimation. The MASSIVE algorithm accounts for model uncertainty by performing Bayesian model averaging over the most promising many-dimensional instrumental variable models, while at the same time employing weaker assumptions regarding the data generating process compared to similar methods.

Content

The data set contains source code implementing the MASSIVE algorithm, which is described in the article titled "MASSIVE: Tractable and Robust Bayesian Learning of Many-Dimensional Instrumental Variable Models" by Ioan Gabriel Bucur, Tom Claassen and Tom Heskes. The data set also contains simulated data necessary for reproducing the figures in the article as well as routines necessary for recreating it. This research is presented in Chapter 5 of the PhD thesis titled "Being Bayesian about Causal Inference" by Ioan Gabriel Bucur. The code is written in the R and C++ programming languages.

Structure

The code is structured on the skeleton of an R package package as follows:

• The folder data contains pre-saved simulated data in .RData format, which we use to recreate the figures from the article. The simulated data can also be reproduced using the reproduce-data.R file from the scripts folder. The simulated data sets are described in R/data.R.

• The folder R contains the R files that implement the MASSIVE algorithm and the routines necessary for reproducing the figures from the article. The main method is implemented in R/MASSIVE.R.

• The folder man contains the documentation for the implemented functions.

• The folder src contains an efficient Rcpp implementation for computing the MASSIVE model posterior, gradient, and Hessian matrix.

• The folder scripts contains the script uai-article-figures.R, which can be run from R to produce the figures in the UAI . The script reproduce-data.R is meant for reproducing the simulated data used in uai-article-figures.R.

• The folder tests contains a few basic unit tests for the R package.

• The folder inst contains additional files used during installation.

  • The subfolder inst/stan contains a Stan file implementing the MASSIVE model, used in reproduce-data.R for comparing priors.
  • The subfolder inst/extdata contains a few data files that are not in .RData format. These contains a list of genetic variants and their associations with BMI, the risk of psoriasis, as well as the effect allele frequencies (EAF).
    • GWAS_BMI.txt corresponds to Table B (Association of BMI genetic instruments with BMI in UK Biobank and HUNT) in the Supplementary Material of (Budu-Aggrey et al., 2019).
    • GWAS_psoriasis.txt corresponds to Table 3 (Association of BMI genetic instruments with psoriasis in UK Biobank and HUNT) in the Supplementary Material of (Budu-Aggrey et al., 2019).
    • GWAS_EAF.txt corresponds to Supplementary Table 4 (Significant loci for BMI at P < 1 × 1e-5 for European sex-combined analysis) from the Supplementary Data of (Locke et al., 2015).

• The top folder also contains the following files:

  • DESCRIPTION is the file describing the R package
  • NAMESPACE is the file specifying the functions provided by the R package
  • LICENSE.md is the file containing the GPL-3 license

Prerequisites

In order to install the software, R must be downloaded and installed. The suggested package R2BGLiMS, necessary for running the competing JAM-MR method, also requires the Java JDK to be installed.

Installation Instructions

Download the software from GitHub with the following command: git clone https://github.com/igbucur/MASSIVE.git. For installing and running the MASSIVE R package, several R package are required. These are specified in the package DESCRIPTION file.

To install the package, open an R instance and run (from the MASSIVE folder):

install.packages('devtools') # required package
devtools::install_deps(".", dependencies = TRUE) # install MASSIVE package dependencies
install.packages(".", repos = NULL, type = 'source') # install MASSIVE

library(MASSIVE) # load the package
help(package = "MASSIVE") # see available functions

Licensing

MASSIVE algorithm - Model Assessment and Stochastic Search for Instrumental Variable Estimation

Copyright (C) 2020 Ioan Gabriel Bucur ioan.gabriel.bucur@gmail.com

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.