EDISON (version 1.1.1)

phase.update: Make a network structure or hyperparameter move.

Description

This function makes a network structure or information sharing hyperparameter move.

Usage

phase.update(Eall, Sall, Ball, Sig2all, X, Y, GLOBvar, HYPERvar, target)

Arguments

Eall

List of changepoints with one entry for each target node. Each entry has length equal to the number of changepoints for that target node.

Sall

Network structure: List of length equal to the number of target nodes, where each list entry is a NumSegs by NumNodes matrix.

Ball

Network structure with regression coefficients: Same as Sall, but with regression coefficients as matrix entries.

Sig2all

Sigma squared.

X

Input response data.

Y

Input target data.

GLOBvar

Global variables used during the MCMC simulation.

HYPERvar

Hyperparameter variables.

target

Current target node.

Value

Returns a list with the following elements:

E

Changepoints for the current target node.

Sall

Network structure (possibly updated).

Ball

Network structure regression coefficients (possibly updated).

Sig2all

Sigma squared.

prior.params

Information sharing prior hyperparameters (possibly updated).

k

Level-2 exponential prior hyperparameter (possibly updated).

move

Move type: 4 for a network structure move, 5 hyperparameter move.

move

Structure Move type: 1 for a network structure move, 2 for a level-1 hyperparameter move, 3 for a level-2 hyperparameter move.

accept

1 if the move has been accepted, 0 otherwise.

References

For more information on network structure moves and information sharing priors, see:

Dondelinger et al. (2012), "Non-homogeneous dynamic Bayesian networks with Bayesian regularization for inferring gene regulatory networks with gradually time-varying structure", Machine Learning.

See Also

make_structure_move