Skip to contents

subtract_networks() subtracts one network's weights from another's, cell by cell, and returns the difference x - y as a netdifference object: a network whose edge weights are the signed differences. It takes any pair of networks on the same nodes (netobjects, cograph_networks or square numeric matrices); for an edge-betweenness difference, subtract two net_edge_betweenness results. cograph::splot(d) or cograph::plot_difference(d) draws the signed difference network, with cograph's colouring and node palette.

The difference is taken between two point estimates and carries no uncertainty, so a large difference on a rarely observed edge can be noise. To see which edges differ beyond sampling variation, compare the two networks with permutation() or bayes_compare(); as_netdifference() turns a bayes_compare() result into the same kind of object, keeping, by default, only the credible differences.

Usage

subtract_networks(x, y)

# S3 method for class 'netdifference'
print(x, max_print = 12L, ...)

Arguments

x, y

The two networks: netobjects, cograph_networks or square numeric matrices. They must have the same number of nodes, with the same labels in the same order, or an error is raised; any other kind of object raises an error too. A matrix without dimnames is labelled V1, V2, ..., so it matches only a network with those labels. The difference is directed when x is a directed netobject; when x is not a netobject, it is directed when x's weights are not symmetric. For the print() method: x is the netdifference object.

max_print

For the print() method: the most edges to list; the number left out is reported. Default 12.

...

For the print() method: ignored.

Value

A netdifference object, which is also a netobject (method "difference"). print() lists the edges present in either network, largest absolute difference first, with both weights and their difference (for an undirected difference, each pair once); cograph::splot() draws it. Beside the usual netobject fields, its weights and difference_matrix both hold x - y, and x and y hold the two weight matrices that were subtracted.

The print() method returns its input invisibly.

Examples

early <- data.frame(
  V1 = c("A","B","A","C"), V2 = c("B","C","B","A"),
  V3 = c("C","A","C","B"))
late <- data.frame(
  V1 = c("B","A","C","B"), V2 = c("C","B","A","C"),
  V3 = c("A","C","B","A"))
a <- build_network(early, method = "relative")
b <- build_network(late, method = "relative")
subtract_networks(a, b)
#> Network difference (x - y): 3 nodes, 0 differing edges
#> Plot: cograph::splot(d) or cograph::plot_difference(d)
#> 
#>  from to x y difference
#>     C  A 1 1          0
#>     A  B 1 1          0
#>     B  C 1 1          0
# edge-betweenness difference:
subtract_networks(net_edge_betweenness(a), net_edge_betweenness(b))
#> Network difference (x - y): 3 nodes, 0 differing edges
#> Plot: cograph::splot(d) or cograph::plot_difference(d)
#> 
#>  from to x y difference
#>     C  A 3 3          0
#>     A  B 3 3          0
#>     B  C 3 3          0