Skip to contents

Network-Based Statistic (NBS)

Source: R/stats-network.R
networkBasedStatistic.Rd

Identifies connected subnetworks of edges that differ between two groups (or two within-subject conditions) of connectivity matrices, with family-wise error control from a permutation null on the largest suprathreshold component (Zalesky, Fornito & Bullmore 2010). Each edge is first tested with a mass-univariate t-test; edges whose statistic exceeds thresh form a suprathreshold graph whose connected components are the candidate subnetworks. Their size (edge count) or mass (summed statistic exceedance, Smith 2009) is compared against the permutation distribution of the largest component to obtain FWER-corrected component p-values.

Usage

networkBasedStatistic(
  mats_group1,
  mats_group2,
  thresh = 3,
  n_perm = 1000L,
  tail = c("both", "right", "left"),
  paired = FALSE,
  component = c("size", "mass"),
  directed = FALSE,
  alpha = 0.05,
  seed = NULL
)

Arguments

mats_group1

A list of n x n connectivity matrices or an n x n x N array (group 1, or condition 1 when paired).

mats_group2

The corresponding matrices for group 2 (or condition 2 for a paired design).

thresh

Primary edge-statistic (t) threshold (default: 3).

n_perm

Number of permutations (default: 1000).

tail

"both", "right" (group1 > group2), or "left".

paired

Logical; TRUE for a within-subject (paired) design, in which case the two inputs are paired condition matrices (default: FALSE).

component

Component measure: "size" (edge count) or "mass" (summed statistic exceedance).

directed

Logical; use all off-diagonal (directed) edges rather than the upper triangle (default: FALSE).

alpha

Significance level for the returned adjacency mask (default: 0.05).

seed

Optional RNG seed for reproducible permutations.

Value

A list with components (a data.frame of component size, mass, and p_value), component_edges (a list of node-pair matrices, one per component), the adjacency mask of edges in significant components, the edge_stats matrix, the suprathreshold mask, the permutation null_distribution, and the settings used.

References

Zalesky, A., Fornito, A., & Bullmore, E. T. (2010). Network-based statistic: identifying differences in brain networks. NeuroImage, 53(4), 1197-1207.

See also

plotNBSnetwork()

Examples

if (FALSE) { # \dontrun{
g1 <- replicate(20, matrix(rnorm(400), 20), simplify = FALSE)
g2 <- replicate(20, matrix(rnorm(400), 20), simplify = FALSE)
nbs <- networkBasedStatistic(g1, g2, thresh = 3, n_perm = 500)
nbs$components
} # }