networktools (version 1.1.0)

impact.NCT: Network Comparison Test for Impact Statistics

Description

This function wraps the function NCT from the NetworkComparisonTest package to provide an explicit test for the significance of node impacts.

Usage

impact.NCT(input, it, gamma, nodes = c("all"), binary.data = FALSE,
  weighted = TRUE, split = c("median", "mean", "forceEqual", "cutEqual",
  "quartiles"), paired = FALSE, AND = TRUE, test.edges = FALSE, edges,
  progressbar = TRUE)

Arguments

input

a matrix or data frame of observations (not a network/edgelist). See included example datasets depression and social.

it

the number of iterations (permutations) in each network comparison test

gamma

the sparsity parameter used in generating networks. Defaults to 0.5 for interval data and 0.25 for binary data

nodes

indicates which nodes should be tested. Can be given as a character string of desired nodes (e.g., c("node1","node2")) or as a numeric vector of column numbers (e.g., c(1,2)).

binary.data

logical. Indicates whether the input data is binary

weighted

logical. Indicates whether resultant networks preserve edge weights or binarize edges.

split

method by which to split network given non-binary data. "median": median split (excluding the median), "mean": mean split, "forceEqual": creates equally sized groups by partitioning random median observations to the smaller group, "cutEqual": creates equally sized groups by deleting random values from the bigger group,"quartile": uses the top and bottom quartile as groups

paired

Logical. Can be TRUE of FALSE to indicate whether the samples are dependent or not. If paired is TRUE, relabeling is performed within each pair of observations. If paired is FALSE, relabeling is not restricted to pairs of observations. Note that, currently, dependent data is assumed to entail one group measured twice.

AND

Logical. Can be TRUE of FALSE to indicate whether the AND-rule or the OR-rule should be used to define the edges in the network. Defaults to TRUE. Only necessary for binary data.

test.edges

Logical. Can be TRUE of FALSE to indicate whether or not differences in individual edges should be tested.

edges

Character or list. When 'all', differences between all individual edges are tested. When provided a list with one or more pairs of indices referring to variables, the provided edges are tested. A Holm-Bonferroni correction is applied to control for multiple testing.

progressbar

Logical. Should the pbar be plotted in order to see the progress of the estimation procedure? Defaults to TRUE.

...

additional optional arguments to be passed to the NCT function internally (paired, AND, test.edges, edges, progressbar)

Value

impact returns a list where each element is an object of class NCT

Details

The NCT method is computationally intensive. It is recommended that users test a subset of nodes at a time using the nodes argument, rather than testing all nodes simultaneously.

In order to be interpreted in a meaningful way, the significance of impact statistics should be explicitly tested.

The NCT function from the NetworkComparisonTest uses a permutation test to determine the significance of structure invariances between two networks. Because impact statistics are mathematically defined as structural invariance between two networks, NCT is an appropriate method to test the significance of impact statistics.

impact.NCT returns an object of class NCT, which includes p-values for invariances.

Examples

# NOT RUN {
out <- impact.NCT(depression[,1:5], it=5, nodes="psychomotor_retardation")
# }
# NOT RUN {
NCT1 <- impact.NCT(depression, it=25, nodes="psychomotor_retardation")
NCT1$psychomotor_retardation$glstrinv.pval
NCT1$psychomotor_retardation$nwinv.pval
## Both significant

NCT2 <- impact.NCT(social, it=25, nodes="Kim", binary.data=TRUE)
NCT2$Kim$glstrinv.pval
NCT2$Kim$nwinv.pval
## Only global strength impact is significant

##Note: for speed, 25 permutations are iterated here; more permutations are necessary in practice
# }