In njetzel/ngc: Methods for Estimating Graphical Granger Causality

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 4.5,
  fig.height = 4.5
)
setwd("/Users/njetzel/Documents/RA/NGC/R")
source("defn.r")
source("simulate.r")
source("array2mat.r")
source("pldag.set.r")
source("ngc.r")
source("grangerLasso.r")
source("grangerTLasso.r")
source("grangerThrLasso.r")
library(glmnet)
library(gglasso)
library(igraph)

NGC implements methods for estimating network Granger causality from longitudinal and time series data. NGC supports network estimation using the Granger lasso, truncating lasso, and adaptively thresholded lasso estimators. These methods are described in Shojaie and Michailidis 2010 and Shojaie et al 2011. NGC also implements group lasso fits with each of these methods. The primary function is ngc, which fits a network estimate. S3 plot and predict functions are also provided.

Requirements

• glmnet
• gglasso
• igraph
• Matrix

Examples

set.seed(1)
p <- 9
len <- 20
d_actual <- 3
d <- 6
n <- 30
sigma <- 0.3
edge <- defn_net(d = d_actual, p = p, n = n)
X <- simulate_data(n, edge, len, error_sd = sigma)
fit1 = ngc(X, d = d, typeIerr = 0.05)
fit2 <- ngc(X, d = d, method = 'threshold', refit = TRUE)
plot(fit1)
plot(fit2)
plot(fit1, ngc.type = "granger")
fit1_pred <- predict(fit1, 2)

References

Shojaie A. and Michailidis G. (2010) Discovering Graphical Granger Causality Using a Truncating Lasso Penalty, Bioinformatics, 26(18): i517-i523

Shojaie A., Basu S. and Michailidis G. (2012) Adaptive Thresholding for Reconstructing Regulatory Networks from Time Course Gene Expression Data, Statistics In Biosciences 4(1): 66-83

Basu S., Shojaie A. and Michailidis G. (2015) Network Granger Causality with Inherent Grouping Structure, Journal of Machine Learning Research (JMLR)

njetzel/ngc documentation built on May 31, 2019, 2:27 p.m.