R: Setup and run the MCMC simulation.

runDBN {EDISON}

R Documentation

Setup and run the MCMC simulation.

Description

This function initialises the variabes for the MCMC simulation, runs the simulation and returns the output.

Usage

runDBN(targetdata, preddata = NULL, q, n, multipleVar = TRUE,
  minPhase = 2, niter = 20000, scaling = TRUE, method = "poisson",
  prior.params = NULL, self.loops = TRUE, k = 15, options = NULL,
  outputFile = ".", fixed.edges = NULL)

Arguments

`targetdata`	Target input data: A matrix of dimensions NumNodes by NumTimePoints.
`preddata`	Optional: Input response data, if different from the target data.
`q`	Number of nodes.
`n`	Number of timepoints.
`multipleVar`	`TRUE` when a specific variance is estimated for each segment, `FALSE` otherwise.
`minPhase`	Minimal segment length.
`niter`	Number of MCMC iterations.
`scaling`	If `TRUE`, scale the input data to mean 0 and standard deviation 1, else leave it unchanged.
`method`	Network structure prior to use: `'poisson'` for a sparse Poisson prior (no information sharing), `'exp_hard'` or `'exp_soft'` for the exponential information sharing prior with hard or soft node coupling, `'bino_hard'` or `'bino_soft'` with hard or soft node coupling.
`prior.params`	Initial hyperparameters for the information sharing prior.
`self.loops`	If `TRUE`, allow self-loops in the network, if `FALSE`, disallow self-loops.
`k`	Initial value for the level-2 hyperparameter of the exponential information sharing prior.
`options`	MCMC options as obtained e.g. by the function `defaultOptions`.
`outputFile`	File where the output of the MCMC simulation should be saved.
`fixed.edges`	Matrix of size NumNodes by NumNodes, with `fixed.edges[i,j]==1\|0` if the edge between nodes i and j is fixed, and -1 otherwise. Defaults to `NULL` (no edges fixed).

Value

A list containing the results of the MCMC simulation: network samples, changepoint samples and hyperparameter samples. For details, see output.

Author(s)

Sophie Lebre

Frank Dondelinger

References

For more information about the MCMC simulations, 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.