FLAG: The flexible and accurate Gaussian graphical model in R

The R package FLAG implements the methods based on the paper Flexible and Accurate Methods for Estimation and Inference of Gaussian Graphical Models with Applications. FLAG aims to estimate precision matrix entries accurately and efficiently, and further quantify the uncertainty of each entry, which allows for better leveraging of the common structure across different groups through meta-analysis.

Installation

For a quick start, you can install the development version of FLAG from GitHub with:

# install.packages("devtools")
devtools::install_github("YangLabHKUST/FLAG")

Example

• This is a basic example which shows you how to solve a common problem:

library(FLAG)

rm(list = ls())
set.seed(1234)
N = 20
P = 10
pi = 0.2

Pre = matrix(sample(c(0.2, 0.4), P*P, replace = T) * rbinom(P*P, 1, pi), nrow = P, ncol = P )
Pre[lower.tri(Pre)] = t(Pre)[lower.tri(Pre)]
diag(Pre) = 1
vals <- eigen(Pre)$values
Sigma = solve(Pre)

Z = MASS::mvrnorm(N, rep(0, P), Sigma)
Z.c = scale(Z, center = T, scale = F)

results = FLAG(Z.c)
results$precision.est

Real data analysis

Gene co-expression in the Human Brain

library(FLAG)

For the detailed experiments, please refer to NumericalResults/RealData/GeneExpression/gene_expression_data_analysis.ipynb

Term association in the university webpages

library(FLAG)

For the detailed experiments, please refer to NumericalResults/RealData/TermAssociation/meta_analysis.ipynb

Price comovement of S&P 100 stocks

library(FLAG)

For the detailed experiments, please refer to NumericalResults/RealData/StockComovement/network_measure.ipynb

Citing our work

If you find the FLAG package or any of the source code in this repository useful for your work, please cite:

Qian, Y., Hu, X., & Yang, C. (2023). Flexible and Accurate Methods for Estimation and Inference of Gaussian Graphical Models with Applications. arXiv e-prints, arXiv-2306. https://doi.org/10.48550/arXiv.2306.17584

Contact

Please feel free to contact Yueqi Qian, Prof. Xianghong Hu, or Prof. Can Yang if any inquiries.