Function Reference

Documentation for the bgms API.

Complete documentation for all exported functions.

bgm() and bgmCompare() return S7 objects (classes bgms and bgmCompare). Standard methods like print(), summary(), and coef() work through S3 dispatch for compatibility with base R generics.

Core Functions

Function Description
bgm() Estimates the main effects and pairwise interactions of a network using a Bayesian framework.
bgmCompare() Estimates whether edge weights and category thresholds differ across networks of two different groups.

Prior Samplers

Draw from the model’s priors before seeing data, for prior visualization and simulation studies.

Function Description
sample_ggm_prior() Prior sampler for the GGM precision matrix (conditional, joint, or hierarchical specification).
sample_graph_prior() Prior sampler for the edge-inclusion indicators and edge-prior hyperparameters.
sample_sbm_prior() Ancestral sampler for the stochastic block model hyperprior (allocations, pair probabilities).

Prior Constructors

Prior specification objects passed to bgm() and bgmCompare() via the interaction_prior, threshold_prior, means_prior, precision_scale_prior, edge_prior, and difference_prior arguments.

Constructor Family Used for
cauchy_prior() Cauchy(0, scale) Interactions, thresholds, means
normal_prior() Normal(0, scale) Interactions (bgm() and bgmCompare() default), thresholds, means
beta_prime_prior() Beta-prime Thresholds (default), interactions
gamma_prior() Gamma(shape, rate or eta) Precision-matrix diagonal
exponential_prior() Exponential(rate or eta) Precision-matrix diagonal (default)
bernoulli_prior() Bernoulli Edge / difference inclusion (default)
beta_bernoulli_prior() Beta-Bernoulli Edge / difference inclusion
sbm_prior() Stochastic Block Model Edge / difference inclusion

Methods

Method Description
print() Minimal console output for a fitted model
summary() Summary statistics for posterior samples
coef() Extract posterior means of model parameters
plot() Draw the fitted model: the edge evidence plot by default, or posterior centrality
plot_edge_posterior() Draw one edge’s posterior against its prior, with the inclusion evidence as a probability wheel
predict() Predict new observations from a fitted model
simulate() Simulate datasets from the posterior distribution
simulate_mrf() Simulate observations from a Markov Random Field (mrfSampler() is the deprecated former name)

Diagnostics

Post-fit checks on a model you have already estimated.

Function Description
verdicts() Read each edge’s inclusion Bayes factor as a presence / undecided / absence verdict, and flag the verdicts a longer run could change.
calibration_check() Reliability diagrams of the model’s conditional predictions, with the consistency band a calibrated model wanders inside.
prior_sensitivity_check() Recover each edge’s inclusion Bayes factor curve across the slab scale and classify which verdicts depend on it.
summarize_zratio_gauge() Trust gauge for the hierarchical graph prior’s fast normalizer approximation.

Extractor Functions

Function Description
extract_arguments() Retrieve the arguments used when fitting a model
extract_main_effects() Extract main-effect parameters: posterior means for bgms fits, posterior samples for bgmCompare fits
extract_pairwise_interactions() Extract posterior samples of partial association parameters
extract_centrality() Evaluate strength centrality on every posterior draw, giving each node’s centrality posterior; summary() and plot() read it
extract_indicators() Extract posterior samples of edge or difference indicators
extract_posterior_inclusion_probabilities() Compute posterior inclusion probabilities, from the Rao-Blackwellized draws by default or from the raw indicator samples
extract_prior_inclusion_probabilities() Compute prior inclusion probabilities, matched in shape to the posterior ones for Bayes factor computation
extract_inclusion_bf() Compute inclusion Bayes factors from the Rao-Blackwellized odds accumulators, on the Bayes factor scale by default or on the log scale with log = TRUE; finite for saturated edges up to the numerical bound
extract_indicator_priors() Retrieve the indicator prior specification (prior type and hyperparameters)
extract_group_params() Extract group-specific parameters from a bgmCompare fit
extract_sbm() Extract stochastic block model assignments and probabilities
extract_ess() Compute effective sample sizes for model parameters; for indicators, the Rao-Blackwellized ESS
extract_rhat() Compute split-R-hat convergence diagnostics
extract_precision() Extract posterior mean precision matrix (GGM / continuous block)
extract_partial_correlations() Extract posterior mean partial correlation matrix
extract_log_odds() Extract posterior mean log-odds matrix (OMRF / discrete block)

Datasets

Dataset N Items Type Description
ADHD 355 18 Binary ADHD symptom ratings for children
Boredom 986 8 Ordinal Boredom proneness scores (7-point Likert)
Wenchuan 362 17 Ordinal PTSD symptom ratings (5-point Likert)