generateNetwork {EDISON}

Generate a random network.

Description

This function generates a random network with changepoints for structure changes, for simulating synthetic data.

Usage

generateNetwork(lambda_2 = 0.45, q = 10, min_phase_length = 1,
  k_bar = 5, l = 10, lambda_3 = 2, spacing = 1, gauss_weights = TRUE,
  same = FALSE, change_method = "sequential", fixed = FALSE, cps = NULL)

Arguments

lambda_2

Average number of parents for each node in the network (parameter for a Poisson distribution).

q

Number of nodes.

min_phase_length

Minimum segment length.

k_bar

Maximum number of changepoints. If fixed=TRUE, this is equal to the number of changepoints.

l

Length of the time series.

lambda_3

Average number of structure changes between two segments (parameter for a Poisson distribution).

spacing

1 if segments are equally spaced, 0 if they are spaced randomly (subject to the constraints of min_phase_length).

gauss_weights

1 if edge weights in the network are drawn from N(0, 1), 0 if they are fixed to be 1.

same

1 if all segments have the same network structure (no changes), 0 otherwise.

change_method

'sequential' if the changes happen sequentially (i.e. changes at segment i are applied to segment i-1), 'hierarchical' if the changes happen with respect to a hypernetwork (i.e. changes at segment i are applied to segment 0).

fixed

T if the changepoint locations are fixed, F if they should be sampled.

cps

Changepoint locations (if they are fixed).

Value

A list with the following elements:

network

The network, a list of length NumSegs, where each element is a NumNodes by NumNodes matrix.

epsilon

The vector of changepoint locations.

k

The number of changepoint.

changes

The number of changes among segments.

Author(s)

Frank Dondelinger

See Also

simulateNetwork

Examples

# Generate random network with default parameters
network = generateNetwork()

# Simulate data using generated network
dataset = simulateNetwork(net=network)

# Generate random network with 4 changepoints and 15 nodes, 
# with changepoints distributed over a timeseries of length 50
network = generateNetwork(l=50, q=15, fixed=TRUE, k_bar=4)

# Simulate data of length 50 using generated network
dataset = simulateNetwork(net=network)