network.AIC: AIC/BIC criterion for network graph

In nem: (Dynamic) Nested Effects Models and Deterministic Effects Propagation Networks to reconstruct phenotypic hierarchies

Description

Calclate AIC/BIC for a given network graph (should be transitively closed). The number of free parameters equals the number of unknown edges in the network graph.

Usage

	network.AIC(network,Pm=NULL,k=length(nodes(network$graph)),verbose=TRUE)

Arguments

network	a nem object (e.g. 'pairwise')
Pm	prior over models (n x n matrix). If NULL, then a matrix of 0s is assumed
k	penalty per parameter in the AIC/BIC calculation. k = 2 for classical AIC
verbose	print out the result

Details

For k = log(n) the BIC (Schwarz criterion) is computed. Usually this function is not called directly but from nemModelSelection

Value

AIC/BIC value

Author(s)

Holger Froehlich

See Also

nemModelSelection

Examples

   data("BoutrosRNAi2002") 
   D = BoutrosRNAiDiscrete[,9:16]
   control = set.default.parameters(unique(colnames(D)), para=c(0.13,0.05))
   res1 <- nem(D, control=control)
   network.AIC(res1)
   control$lambda=100 # enforce sparsity
   res2 <- nem(D,control=control)
   network.AIC(res2)

nem documentation built on Oct. 31, 2019, 2:12 a.m.