Skip to contents

Prunes a network represented by a tna object by removing edges based on a specified threshold, lowest percent of non-zero edge weights, the disparity filter algorithm (Serrano et al., 2009), or bootstrap significance.

Usage

prune(x, ...)

# S3 method for class 'tna'
prune(
  x,
  method = "threshold",
  threshold = 0.1,
  lowest = 0.05,
  level = 0.5,
  boot = NULL,
  ...
)

# S3 method for class 'group_tna'
prune(x, ...)

Arguments

x

An object of class tna or group_tna.

...

Arguments passed to bootstrap() when using method = "bootstrap" and when a tna_bootstrap is not supplied. Method-specific arguments supplied for another method are ignored with a warning instead of being silently discarded.

method

A character string describing the pruning method. The available options are "threshold", "lowest", "bootstrap" and "disparity", corresponding to the methods listed in Details. The default is "threshold".

threshold

A numeric value specifying the edge weight threshold. Edges with weights below or equal to this threshold will be considered for removal. The default is 0.1.

lowest

A numeric value specifying the lowest percentage of non-zero edges. This percentage of edges with the lowest weights will be considered for removal. The default is 0.05.

level

A numeric value representing the significance level for the disparity filter. Defaults to 0.5.

boot

A tna_bootstrap object to be used for pruning with method "bootstrap". Supplying a non-NULL value selects bootstrap pruning even if method is omitted. A warning is issued if it conflicts with an explicitly supplied method.

Value

A pruned tna or group_tna object: the removed edges have weight zero in weights, and the pruning is recorded in the attribute pruning. Details on the pruning can be viewed with pruning_details(). The original model can be restored with deprune().

Details

The available methods are:

  • "threshold": Removes the edges whose weight is at most threshold.

  • "lowest": Removes the edges whose weight is at most the lowest quantile of the positive edge weights.

  • "disparity": Keeps the edges that are significant at level level under the disparity filter of Serrano et al. (2009), either among the outgoing edges of their source state or among the incoming edges of their target state. Self-loops are always removed by this method.

  • "bootstrap": Keeps the edges that are significant in bootstrap() (the model element of a tna_bootstrap object).

With "threshold" and "lowest", an edge is only removed when the network stays weakly connected without it, so some edges below the cut-off can be retained. The "disparity" and "bootstrap" methods do not check the connectivity.

References

Serrano, M. A., Boguna, M., & Vespignani, A. (2009). Extracting the multiscale backbone of complex weighted networks. Proceedings of the National Academy of Sciences, 106(16), 6483–6488. doi:10.1073/pnas.0808904106

Examples

model <- tna(group_regulation)
pruned_threshold <- prune(model, method = "threshold", threshold = 0.1)
pruned_percentile <- prune(model, method = "lowest", lowest = 0.05)
pruned_disparity <- prune(model, method = "disparity", level = 0.5)