richardli / EMGS

EM for Gaussian graphical model selection

EMGS

This repository contains the R package EMGS (currently still developer's version), described in

Zehang R Li and Tyler H McCormick. An Expectation Conditional Maximization approach for Gaussian graphical models, 2018 arXiv

Using the package

You can install the package with devtools

install_github("richardli/EMGS", subdir = "EMGS")
library(EMGS)

First naive example

This example demonstrates the bias reduction of EMGS. It creates a plot (illustraion-emgs.pdf) under the figures/ directory.

setwd("codes/")
source("example1.R")

Second example

This example demonstrates the informative priors. It creates a plot (structure.pdf) under the figures/ directory.

source("example2.R")

Third example

This example demonstrates the missing data imputation for Burke Gilman Trail example. It creates a plot (burke.pdf) under the figures/ directory. For quicker illustration, this example models only two blocks by combining the pedestrians and bikes from both directions instead of modeling four blocks.

source("example3.R")

Simulation studies

Running the full version of the simulation as described in the paper takes a long time and is recommended to be implemented on a cluster. Below is a low dimensional example.

source("simulation-sim.R")
source("simulation-example.R")