tsnetwork: Time Series Chain Graphical Models for mixed...

In tsnetwork: Constructing Dynamic Networks for Time-Series Data

Description

Reconstructing instantaneous (undirected) and dynamic (directed) networks from repeated multivariate mixed discrete-continuous or ordinal time series data. This function computes sparse an autoregressive coefficient and a precision matrices for time series chain graphical models.

Usage

 tsnetwork(dat, lower, upper, penalty= "scad", n_lam1= NULL, 
      lam1_ratio= NULL, n_lam2= NULL, lam2_ratio= NULL, em_tol= NULL,
      em_iter= NULL, iter_Mstep= NULL, pen_diag_gamma= FALSE,
      ncores = 1)

Arguments

dat	Longitudinal data format
lower	Lower boundry of the data. Can be cimouted internally. Deafult is NULL.
upper	Upper boundry of the data. Can be cimouted internally. Deafult is NULL.
penalty	This specifies the type of penalty function to be used. SCAD penalty function is applied if penalty = "scad" and GLASSO is applied if penalty = "lasso"
n_lam1	The number of regularization parameters for the instantaneous interactions.
lam1_ratio	Determines the sequence of lam1.
n_lam2	The number of regularization parameters for the dynamics interactions.
lam2_ratio	Determines the sequence of lam2.
em_tol	A value to meet the convergence criteria of the EM algorithm. Default value is 0.01
em_iter	The number of EM iterations. The default value is 10.
iter_Mstep	The number of iterations in the M-step to guarantee the convergence. The default value is 5.
pen_diag_gamma	Penalazing the diagonal elements of the autoregressive matrix.
ncores	The number of cores to use for the calculations.

Author(s)

Pariya Behrouzi
Maintainer: Pariya Behrouzi <[e-mail]>

References

Pariya Behrouzi, Fentaw Abegaz and Ernst Wit (2018). Dynamic Chain Graph Models for Ordinal Time Series Data. Arxiv. 14, 3: 586-599.

Fentaw Abegaz and Ernst Wit (2013). Sparse time series chain graphical models for reconstructing genetic networks. Biostatistics. 14, 3: 586-599.

Examples

simulate <- gen.sim(t = 3, n = 2, p = 3, k = 3, network = "scale-free") 
sim.dat <- simulate$dat
out <- tsnetwork(dat =sim.dat, lower= NULL, upper= NULL, penalty= "lasso",
                 n_lam1= 1, lam1_ratio= NULL, n_lam2= 1, lam2_ratio= NULL, em_tol = NULL,
                 em_iter= 1, iter_Mstep = 1, pen_diag_gamma= FALSE, ncores = 1)

# Estimated sparse precision (undirected) and autoregression (directed) matrices
undirected <- out$theta[ , , 1, 1]
directed <- out$gamma[ , , 1, 1]

oldpar <- par(no.readonly =TRUE)
par(mfrow=c(1,2))
plotG(undirected, mod="undirected", main= "Estimated precision matrix", label=TRUE)     
plotG(directed, mod="directed", main ="Estimated autoregression coef. matrix", label=TRUE) 
par(oldpar)

tsnetwork documentation built on March 26, 2020, 6:45 p.m.