estnet_bayes: Determine the posterior distributions of the network...

In ghosoya/ludwig: Simulating and Estimating Undirected Binary Networks Using the Pseudolikelihood Method

Description

This function allows for estimating an undirected graphical model using MCMC via JAGS (Just another Gibb's sampler).

Usage

estnet_bayes(data, n_chains = 2, n_iter = 1000, n_burnin = 1000,
  n_adapt = 1000, n_thin = 4)

Arguments

data	a binary input matrix
n_chains	the number of chains used
n_iter	the number of MCMC iterations
n_burnin	the number of burn-in iterations
n_adapt	the number of adaption iterations
n_thin	the thinning interval

Value

The function returns a list of class estnet_bayes, containing...

• the objects returned by the function create_matrix

• s_coda the raw coda samples

• fitted expected values of the nodes given the posterior means of the network parameters

• entropy_rate the entropy rate based on the network parameters' posterior means

• logL the log pseudo likelihood

• deviance the deviance

• merged chains the merged MCMC chains

• used_samples the actual number of samples to assess the posterior distributions

• posterior_means the parameter's posterior means

• coef the parameter's posterior means

• posterior_sd the parameter's posterior standard deviations

• weights a matrix of the network parameter's posterior means

• weights_sd a matrix of the network parameter's posterior standard deviations

• weights the estimated weights

• thresholds posterior means of the threshold parameters

• thresholds_sd posterior standard deviations of the threshold parameters

• n_chains number of MCMC chains

• thinning thinning interval

• burnin burnin iterations

• n_iter number of iterations

• actual actual iterations used to compute the posterior statistics

• time the time used for the estimation process

• jags the JAGS model object

Examples

# Simulates two negatively coupled nodes
# using random updating
simnet(2, c(1,1,-2), 1000, 1000)

ghosoya/ludwig documentation built on Sept. 20, 2020, 3:48 p.m.