Skip to contents

net_aggregate_weights() reduces a vector of edge weights to one number: their sum, mean, median, maximum, minimum, product, density or geometric mean. Missing and zero weights are removed first, so every method works on the edges that are present. cluster_summary() and build_mcml() use it to combine the edges between two clusters into one cluster-level weight.

Usage

net_aggregate_weights(w, method = "sum", n_possible = NULL)

Arguments

w

The edge weights, a numeric vector. NA is allowed and removed with the zeros; an infinite value raises an error.

method

How to combine the weights: "sum" (default), "mean", "median", "max", "min", "prod", "density" or "geomean". A single value; any other raises an error. "density" divides the sum by n_possible, and "mean" by the number of non-zero weights, so the two agree when every possible edge is present, and "density" is the smaller when some are zero or missing. "geomean" is the geometric mean of the positive weights; negative weights are left out, and the result is 0 when none is positive.

n_possible

For method = "density": the number of possible edges, such as the block size n_i * n_j of two clusters, a single finite number. Default NULL, under which (as with a value of 0 or less) "density" returns the mean of the non-zero weights. Other methods ignore it.

Value

A single number: the chosen aggregate of the non-zero, non-missing weights, or 0 when none remain.

Details

"sum", "prod", "min", "max", "mean" and "median" mean what they mean in igraph's edge.attr.comb argument.

Examples

w <- c(0.5, 0.8, 0.3, 0.9)
net_aggregate_weights(w, "sum")   # 2.5
#> [1] 2.5
net_aggregate_weights(w, "mean")  # 0.625
#> [1] 0.625
net_aggregate_weights(w, "max")   # 0.9
#> [1] 0.9
net_aggregate_weights(w, "density", n_possible = 9)  # 2.5 / 9
#> [1] 0.2777778