gain.functions: MFCF Gain Functions

In NetworkToolbox: Methods and Measures for Brain, Cognitive, and Psychometric Network Analysis

Description

These functions maximize a gain criterion for adding a node to a clique (and the larger network). The flexibility of MFCF allows for any multivariate function to be used as a scoring function.

• "logLik"

  The log determinant of the matrix restricted to the separator minus the log determinant of the matrix restricted to the clique.

• "logLik.val"

  "logLik" with a further validation based on the likelihood ratio. If the increase in gain is not significant the routine stops adding nodes to the separator.

• "rSquared.val"

  The R squared from the regression of the node against the clique. Only the clique nodes with a regression coefficient significantly different from zero are added to the separator / new clique. The gain is different from zero only if the F-values is significant, It assumed that the data matrix is a dataset of realizations (i.e., p variables and n observations).

Usage

"logLik"
gfcnv_logdet(data, clique_id, cl, excl_nodes, ctreeControl)

"logLik.val"
gfcnv_logdet_val(data, clique_id, cl, excl_nodes, ctreeControl)

"rSquared.val"
gdcnv_lmfit(data, clique_id, cl, excl_nodes, ctreeControl)

Arguments

data	Matrix or data frame. Can be a dataset or a correlation matrix
clique_id	Numeric. Number corresponding to clique to add another node to
cl	List. List of cliques already assembled in the network
excl_nodes	Numeric vector. A vector of numbers corresponding to nodes not already included in the network
ctreeControl	List (length = 5). A list containing several parameters for controlling the clique tree sizes: min_size Numeric. Minimum number of nodes allowed per clique. Defaults to 1 max_size Numeric. Maximum number of nodes allowed per clique. Defaults to 8 pval Numeric. p-value used to determine cut-offs for nodes to include in a clique. Defaults to .05 pen Numeric. Multiplies the number of edges added to penalize complex models. Similar to the penalty term in AIC drop_sep Boolean. This parameter influences the MFCF only. If TRUE any separator can be used only once, as in the TMFG. use_returns Boolean. Only used in rSquared.val. If set to TRUE the regression is performed on log-returns. Defaults to FALSE

Value

Returns the value with the maximum gain

Author(s)

Guido Previde Massara <[e-mail]> and Alexander Christensen <[e-mail]>

References

Massara, G. P. & Aste, T. (2019). Learning clique forests. ArXiv.

NetworkToolbox documentation built on May 28, 2021, 5:11 p.m.