LOGGLE (LOcal Group Graphical Lasso Estimation)

Description

The R package loggle provides a set of methods that learn time-varying graphical models based on data measured over a temporal grid. loggle is motivated by the needs to describe and understand evolving interacting relationships among a set of random variables in many real applications, for instance, the gene regulatory networks over the course of organismal development, and the dynamic relationships between individuals in a community over a few years. loggle estimates time-varying graphical models under the assumption that the graph topology changes gradually over time.

loggle has been applied to S&P 500 stock price dataset, where the interacting relationships among stocks and among industrial sectors in a time period that covers the recent global financial crisis can be revealed. Detailed description of S&P 500 stock price dataset is in ?stockdata.

For more details on estimating time-varying graphical models and the package, please refer to: Estimating Time-Varying Graphical Models https://arxiv.org/abs/1804.03811.

Dependencies

Please make sure to install the following package dependencies before using R package loggle. R with version later than 3.0.2 is needed.

install.packages(c("Matrix", "doParallel", "igraph", "glasso", "sm"))

Installation

The R package loggle can be installed from source files in the GitHub repository (R package devtools is needed):

library(devtools)
install_github(repo="jlyang1990/loggle")

Main functions

• loggle: learn time-varying graphical models for a given set of tuning parameters.
• loggle.cv: conduct model selection via cross validation for learning time-varying graphical models.
• loggle.cv.select: conduct model selection for time-varying graphical models based on cross validation results from loggle.cv.
• loggle.cv.vote: learn time-varying graphical models for a given set of tuning parameters via cv.vote.
• loggle.refit: conduct model refitting given learned time-varying graph structures.

Contact

Please report any bugs to jlyang@ucdavis.edu.