Stepwise Gaussian graphical model selection (ggmModSelect)

Selects a GGM by extended-BIC model search over edge sets generated from the glasso path, refitting the unregularized maximum-likelihood precision on each candidate graph, with an optional stepwise add/drop search. Unlike the graphical lasso, retained edges are not shrunk. Equivalent in purpose to qgraph::ggmModSelect(), but pure base R and self-certified via ggm_support_kkt().

Usage

ggm_modselect(
  data = NULL,
  cor_matrix = NULL,
  n = NULL,
  gamma = 0,
  stepwise = TRUE,
  nlambda = 100L,
  lambda_min_ratio = 0.01,
  threshold = 0,
  cor_method = c("pearson", "spearman", "kendall", "auto"),
  na_method = c("pairwise", "listwise"),
  native = TRUE,
  labels = NULL
)

Arguments

data

Numeric data frame or matrix (rows = observations). Optional if cor_matrix is supplied.

cor_matrix

Optional correlation matrix; if given, n is required.

n

Sample size (required when cor_matrix is supplied).

gamma

EBIC hyperparameter. Default 0, matching qgraph::ggmModSelect(): because the selected graph is refit with the unregularized MLE (no edge shrinkage), the extra EBIC penalty is not needed, so gamma 0 (plain BIC) is the method's intended setting. (The regularized GGMs ebic_glasso() and huge_network() keep gamma 0.5.)

stepwise

If TRUE (default), refine the best glasso-path graph by a greedy single-edge add/drop search.

nlambda

Number of glasso penalties scanned for candidate graphs.

lambda_min_ratio

Smallest penalty as a fraction of the largest.

threshold

Partial correlations with absolute value below this are zeroed. Default 0.

cor_method

Correlation used when data is supplied: "pearson" (default), "spearman", "kendall", or "auto" (polychoric/polyserial, the qgraph/bootnet default for ordinal items). See cor_auto().

na_method

Missing-data handling when data is supplied: "pairwise" (default) or "listwise". See ebic_glasso().

native

Solver switch for generating candidate supports: TRUE (default) uses the pure-R solver; FALSE delegates to the glasso Fortran package (in Suggests). The reported precision is the unregularized refit either way. See ebic_glasso().

labels

Optional node labels.

Value

A psychnet object whose $weights is the partial-correlation matrix, with $precision, $support (the selected graph), $gamma, $ebic, $cor_matrix, and $kkt.

Examples

S <- 0.5^abs(outer(1:6, 1:6, "-"))
ggm_modselect(cor_matrix = S, n = 250)
#> <psychnet> ggm network
#>   nodes: 6   edges: 5   (undirected)
#>   optimality (KKT residual): 2.22e-16