Ising network for binary data — ising

Estimates an Ising model by nodewise L1-penalized logistic regression with EBIC selection, combined by the AND (default) or OR rule. Equivalent in purpose to IsingFit::IsingFit(), but pure base R and self-certified: each node's regression reports its stationarity (KKT) residual (see glm_lasso_kkt()).

Usage

ising_fit(
  data,
  gamma = 0.25,
  rule = c("AND", "OR"),
  nlambda = 100L,
  lambda_min_ratio = 0.01,
  min_sum = NULL,
  weights = NULL,
  na_method = c("pairwise", "listwise"),
  native = TRUE,
  labels = NULL
)

Arguments

data: Binary (0/1) data frame or matrix (rows = observations).
gamma: EBIC hyperparameter. Default 0.25.
rule: Edge-combination rule: "AND" (default) or "OR".
nlambda: Number of penalties per nodewise path. Default 100.
lambda_min_ratio: Smallest penalty as a fraction of the largest.
min_sum: Minimum row sum-score (number of endorsed items); rows below it are dropped before fitting. NULL (default) keeps every row.
weights: Optional non-negative observation weights, one per retained row. NULL (default) is unweighted.
na_method: Missing-data handling: "pairwise" (default) single-imputes each column over its observed values (mode for binary), keeping the full sample; "listwise" drops incomplete rows. Identical for complete data.
native: Solver switch. TRUE (default) uses psychnet's own pure-R, dependency-free, self-certified L1 logistic path (KKT ~1e-9). FALSE delegates each per-node fit to the glmnet package with the IsingFit EBIC path, so the returned $weights/$thresholds byte-match IsingFit::IsingFit() (to ~1e-16) at the cost of glmnet's looser self-certificate. native = FALSE needs the optional glmnet package (Suggests); weights/min_sum are supported with native = TRUE only.
labels: Optional node labels.

Value

A psychnet object whose $weights is the symmetric weight matrix, with $thresholds (node intercepts) and $kkt (the worst nodewise stationarity residual).

Examples

set.seed(1)
z <- matrix(stats::rnorm(400 * 2), 400, 2)
x <- cbind(z[, 1], z[, 1], z[, 2], z[, 2]) + matrix(stats::rnorm(400 * 4), 400)
b <- (x > 0) * 1L
colnames(b) <- paste0("V", 1:4)
ising_fit(b)
#> <psychnet> ising network
#>   nodes: 4   edges: 2   (undirected)
#>   optimality (KKT residual): 3.77e-14