expectedInf function

Expected Influence

Calculates the one-step and two-step expected influence of each node.

expectedInf(network, step = c("both", 1, 2), directed = FALSE)

Arguments

• network: an object of type qgraph, igraph, or an adjacency matrix representing a network. Adjacency matrices should be complete (e.g., not only upper or lower half)
• step: compute 1-step expected influence, 2-step expected influence, or both
• directed: logical. Specifies if edges are directed, defaults to FALSE

Details

When a network contains both positive and negative edges, traditional centrality measures such as strength centrality may not accurately predict node influence on the network. Robinaugh, Millner, & McNally (2016) showed that in these cases, expected influence is a more appropriate measure.

One-step expected influence is defined as the sum of all edges extending from a given node (where the sign of each edge is maintained).

Two-step expected influence, as the name implies, measures connectivity up to two edges away from the node. It is defined as the sum of the (weighted) expected influences of each node connected to the initial node plus the one-step expected influence of the initial node. Weights are determined by the edge strength between the initial node and each "second step" node.

See citations in the references section for further details.

Examples

out1 <- expectedInf(cor(depression[,1:5]))

out1$step1
out1$step2
plot(out1)
plot(out1, order="value", zscore=TRUE)

igraph_obj <- igraph::graph_from_adjacency_matrix(cor(depression))
out_igraph <- expectedInf(igraph_obj)

qgraph_obj <- qgraph::qgraph(cor(depression), DoNotPlot=TRUE)
out_qgraph <- expectedInf(qgraph_obj)

References

Robinaugh, D. J., Millner, A. J., & McNally, R. J. (2016). Identifying highly influential nodes in the complicated grief network. Journal of abnormal psychology, 125, 747.