Package 'psychnets'

Description

The canonical (str-visible) fields are the lean netobject set; this method adds virtual aliases so older/external accessors keep working without storing redundant fields.

Usage

## S3 method for class 'psychnet'
x$name

Arguments

Value

The requested field, or NULL if neither a canonical field nor a known alias.

Description

Tidy edge list for a psychnet network

Usage

## S3 method for class 'psychnet'
as.data.frame(x, row.names = NULL, optional = FALSE, ..., include_zero = FALSE)

Arguments

Value

A one-row-per-edge data.frame with columns from, to, weight.

Description

Tidy a network bootstrap

Usage

## S3 method for class 'psychnet_bootstrap'
as.data.frame(x, ...)

Arguments

Value

The tidy ⁠$edges⁠ data frame (one row per edge, with its percentile interval, inclusion proportion, and significant flag).

Description

Every regularized or constrained estimator in psychnet self-certifies: it reports how far the returned network sits from the unique optimum of its own convex objective (a KKT / stationarity residual), or – for the structural methods – whether the graph satisfies the identity that defines it. This verb returns that certificate as a tidy one-row data.frame, so correctness is read the same way for every method.

Usage

certificate(x, tol = 1e-06)

Arguments

Details

The residual is near machine zero for a correctly solved problem. cor and pcor have no optimization to certify and report NA.

Value

A one-row data.frame with columns method, certificate (the residual; smaller is better), kind ("kkt" for the optimization certificates, "structural" for TMFG/relimp, "none" for cor/pcor), and certified (logical: residual at or below tol).

Examples

S <- 0.4^abs(outer(1:6, 1:6, "-"))
certificate(ebic_glasso(cor_matrix = S, n = 250))
certificate(tmfg_network(cor_matrix = S))

Description

Extracts the effective pairwise network implied by a moderated MGM (mgm_fit() with moderators) at a given value of the moderator, mirroring mgm::condition(): it applies the AND-rule pre-filter, absorbs the moderator value into the main-effect coefficients, and re-aggregates the pairwise edges.

Usage

condition(object, value, rule = NULL)

Arguments

Value

A psychnet network object (the moderator node carries no edges).

Examples

set.seed(1)
x1 <- stats::rnorm(400); x2 <- stats::rnorm(400)
mod <- rep(0:1, each = 200)
y <- x1 * (mod == 1) + stats::rnorm(400)   # x1-y edge only when mod == 1
d <- data.frame(x1 = x1, x2 = x2, y = y, mod = mod)
fit <- mgm_fit(d, types = c("g", "g", "g", "c"), moderators = 4)
condition(fit, value = 1)

Description

Detects ordinal variables (integer-valued with at most ordinal_max_levels levels) and returns the correlation matrix using a polychoric correlation for ordinal-ordinal pairs, a polyserial correlation for ordinal-continuous pairs, and Pearson otherwise, projected to the nearest positive-definite matrix. The base-R counterpart of qgraph::cor_auto(); this is the correlation bootnet/qgraph use by default for Likert data.

Usage

cor_auto(data, ordinal_max_levels = 7L, na_method = c("pairwise", "listwise"))

Arguments

Value

A correlation matrix with the variable names as dimnames.

Examples

set.seed(1)
z <- matrix(stats::rnorm(300 * 4), 300, 4) %*% chol(0.5^abs(outer(1:4, 1:4, "-")))
x <- apply(z, 2, function(col) as.integer(cut(col, 5)))   # 5-level Likert
cor_auto(x)

Description

Marginal (zero-order) association network: the Pearson correlation matrix with the diagonal removed. Equivalent to bootnet's "cor" default.

Usage

cor_network(
  data = NULL,
  cor_matrix = NULL,
  n = NULL,
  cor_method = c("pearson", "spearman", "kendall", "auto"),
  threshold = 0,
  alpha = NULL,
  adjust = "none",
  na_method = c("pairwise", "listwise"),
  labels = NULL
)

Arguments

Value

A psychnet object whose ⁠$weights⁠ is the thresholded correlation matrix, with ⁠$cor_matrix⁠, ⁠$n_eff⁠, ⁠$na_method⁠ (and ⁠$p_values⁠ when alpha is used).

Examples

x <- matrix(stats::rnorm(200 * 4), 200, 4)
cor_network(x)
cor_network(x, alpha = 0.05, adjust = "BH")

Description

Splits each column of a numeric matrix or data frame into a binary 0/1 variable. This is the usual preprocessing step before fitting an Ising network (ising_fit(), ising_sampler()) to Likert or other ordinal/continuous data, which require binary input.

Usage

dichotomize(data, method = c("median", "mean", "rank"))

Arguments

Value

An integer matrix of 0/1 values with the same dimensions and dimnames as data.

Examples

b <- dichotomize(SRL_GPT, method = "median")
table(b)                          # values are 0/1 only

Description

Tests, within a single network, whether two edge weights or two node centralities differ. For every pair it forms the per-resample difference from the stored bootstrap draws, takes the percentile interval of that difference, and flags the pair significant when the interval excludes zero; it also reports the two-sided bootstrap p-value (Epskamp, Borsboom & Fried 2018). This is the within-network counterpart to the edge accuracy intervals reported by net_boot().

Usage

difference_test(boot, type = "edge", ci = NULL, p_adjust = "none")

Arguments

Value

A tidy data frame, one row per pair, with item1, item2, the two observed values, their observed difference, the percentile interval of the bootstrap difference (lower, upper), the two-sided p_value, and a logical significant.

Examples

set.seed(1)
x <- matrix(stats::rnorm(150 * 5), 150, 5) %*% chol(0.4^abs(outer(1:5, 1:5, "-")))
colnames(x) <- paste0("V", 1:5)
bs <- net_boot(x, n_boot = 100)
difference_test(bs, type = "strength")

Description

Selects an L1 penalty by the extended BIC (Foygel & Drton 2010) over a log-spaced path, then refits the chosen penalty to machine precision so the returned network is the certified global optimum of the convex objective. Equivalent in purpose to qgraph::EBICglasso() / bootnet's "EBICglasso" default, but pure base R and self-certified (see glasso_kkt()).

Usage

ebic_glasso(
  data = NULL,
  cor_matrix = NULL,
  n = NULL,
  gamma = 0.5,
  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

Value

A psychnet object whose ⁠$weights⁠ is the partial-correlation matrix, with ⁠$precision⁠, ⁠$lambda⁠, ⁠$gamma⁠, ⁠$cor_matrix⁠, ⁠$ebic⁠, ⁠$native⁠, and ⁠$kkt⁠ (the stationarity residual of the returned network).

Examples

S <- 0.4^abs(outer(1:6, 1:6, "-"))
fit <- ebic_glasso(cor_matrix = S, n = 250)
fit
as.data.frame(fit)

Description

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

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)

Description

Certificate for an unregularized Gaussian graphical model whose precision is constrained to a fixed graph (the estimator behind ggm_modselect() and logo_network()). The maximum-likelihood / maximum-entropy conditions for a Gaussian Markov random field on a graph $G$ are exact: $W_{ij} = S_{ij}$ for every $(i,j)$ on the graph and on the diagonal ($W = \Theta^{-1}$), and $\Theta_{ij} = 0$ for every $(i,j)$ not on the graph. A near-zero return certifies the constrained optimum with no reference solver.

Usage

ggm_support_kkt(theta, cor_matrix, support, active_tol = 1e-08)

Arguments

Value

Maximum absolute stationarity violation (scalar); 0 = exact optimum.

Examples

S <- 0.4^abs(outer(1:6, 1:6, "-"))
fit <- ggm_modselect(cor_matrix = S, n = 250)
ggm_support_kkt(fit$precision, S, fit$support)

Description

A dependency-free correctness certificate for a fitted Gaussian graphical model. For the convex objective

$\min_{\Theta \succ 0} -\log\det\Theta + \mathrm{tr}(S\Theta) + \rho \sum_{i \neq j} |\Theta_{ij}|$

(off-diagonal penalty), let $W = \Theta^{-1}$. The subgradient optimality conditions are $W_{ii} = S_{ii}$; $W_{ij} - S_{ij} = \rho\,\mathrm{sign}(\Theta_{ij})$ where $\Theta_{ij} \neq 0$; and $|W_{ij} - S_{ij}| \le \rho$ otherwise. By strict convexity, a precision matrix with zero violation is the unique global optimum, so a near-zero return certifies correctness independently of any reference solver.

Usage

glasso_kkt(theta, cor_matrix, rho, active_tol = 1e-08)

Arguments

Value

Maximum absolute stationarity violation (scalar); 0 = exact optimum.

Examples

S <- 0.5^abs(outer(1:5, 1:5, "-"))
fit <- ebic_glasso(cor_matrix = S, n = 200)
glasso_kkt(fit$precision, S, fit$lambda)

Description

Dependency-free correctness certificate for a nodewise lasso, analogous to glasso_kkt() for the graphical lasso. With standardized predictors X and fitted mean mu (identity link for gaussian, logistic for binomial), the subgradient conditions are $n^{-1} X_j^\top (y - \mu) = \lambda\, \mathrm{sign}(\beta_j)$ for active coordinates and $|n^{-1} X_j^\top (y - \mu)| \le \lambda$ otherwise. Near-zero certifies the penalized-likelihood optimum.

Usage

glm_lasso_kkt(
  X,
  y,
  b0,
  beta,
  lambda,
  family = "gaussian",
  weights = NULL,
  active_tol = 1e-08
)

Arguments

Value

Maximum absolute stationarity violation (scalar). Near-zero certifies the fit is at the penalized-likelihood optimum.

Examples

set.seed(1)
x <- scale(matrix(stats::rnorm(200 * 3), 200, 3))
y <- as.numeric(x %*% c(0.5, 0, -0.3) + stats::rnorm(200))
fit <- stats::lm.fit(cbind(1, x), y)
glm_lasso_kkt(x, y, fit$coefficients[1], fit$coefficients[-1], lambda = 0)

Description

Estimates a Gaussian graphical model after a rank-based nonparanormal transform that relaxes the multivariate-normal assumption, then selects the L1 penalty by EBIC and refits to the certified optimum. Equivalent in purpose to huge::huge() (nonparanormal) / bootnet's "huge" default, but pure base R and self-certified via glasso_kkt() on the transformed correlation.

Usage

huge_network(
  data = NULL,
  cor_matrix = NULL,
  n = NULL,
  npn = c("shrinkage", "truncation", "skeptic"),
  gamma = 0.5,
  nlambda = 100L,
  lambda_min_ratio = 0.01,
  threshold = 0,
  na_method = c("pairwise", "listwise"),
  native = TRUE,
  labels = NULL
)

Arguments

Value

A psychnet object whose ⁠$weights⁠ is the partial-correlation matrix, with ⁠$precision⁠, ⁠$lambda⁠, ⁠$gamma⁠, ⁠$cor_matrix⁠ (the transformed correlation), ⁠$npn⁠, ⁠$ebic⁠, and ⁠$kkt⁠.

Examples

set.seed(1)
x <- matrix(stats::rnorm(300 * 5), 300, 5)
x <- exp(x %*% chol(0.4^abs(outer(1:5, 1:5, "-"))))   # break normality
huge_network(x)

Description

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

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)

Description

Estimates an Ising model by unpenalized nodewise logistic regression, with optional Wald p-value edge pruning, combined by the AND (default) or OR rule. The unregularized counterpart of ising_fit(); self-certified by the maximum-likelihood score residual (see glm_lasso_kkt() at lambda = 0).

Usage

ising_sampler(
  data,
  rule = c("AND", "OR"),
  alpha = NULL,
  adjust = "none",
  min_sum = NULL,
  weights = NULL,
  na_method = c("pairwise", "listwise"),
  labels = NULL
)

Arguments

Value

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

Examples

set.seed(1)
z <- matrix(stats::rnorm(500 * 2), 500, 2)
x <- cbind(z[, 1], z[, 1], z[, 2], z[, 2]) + matrix(stats::rnorm(500 * 4), 500)
b <- (x > 0) * 1L
colnames(b) <- paste0("V", 1:4)
ising_sampler(b)

Description

By Shapley efficiency, the importance shares a node receives from its predictors sum exactly to that node's full-model R-squared. Returns the maximum absolute deviation from that identity; near zero certifies the decomposition.

Usage

lmg_certificate(x)

Arguments

Value

Maximum absolute deviation of incoming-share sums from the per-node R-squared (scalar); 0 = exact decomposition.

Examples

S <- 0.4^abs(outer(1:5, 1:5, "-"))
lmg_certificate(relimp_network(cor_matrix = S))

Description

Estimates a sparse Gaussian graphical model whose conditional-independence structure is the chordal TMFG: the precision is the closed-form Gaussian Markov random field on that graph (Barfuss et al. 2016). Equivalent in purpose to NetworkToolbox::LoGo() / bootnet's "LoGo" default, pure base R and self-certified via ggm_support_kkt() (the precision reproduces S exactly on the TMFG support).

Usage

logo_network(
  data = NULL,
  cor_matrix = NULL,
  n = NULL,
  cor_method = c("pearson", "spearman", "kendall", "auto"),
  threshold = 0,
  na_method = c("pairwise", "listwise"),
  labels = NULL
)

Arguments

Value

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

Examples

set.seed(1)
x <- matrix(stats::rnorm(300 * 6), 300, 6)
logo_network(x)

Description

Estimates a mixed graphical model by nodewise L1-penalized regression – a gaussian (linear) lasso for continuous nodes and a logistic lasso for binary nodes – with per-node EBIC selection, combined by the AND rule. Equivalent in purpose to mgm::mgm(), but pure base R and self-certified: each node's regression reports its stationarity (KKT) residual (see glm_lasso_kkt()).

Usage

mgm_fit(
  data,
  gamma = 0.25,
  types = NULL,
  nlambda = 100L,
  lambda_min_ratio = 0.01,
  threshold = c("LW", "HW", "none"),
  rule = c("AND", "OR"),
  moderators = NULL,
  weights = NULL,
  na_method = c("pairwise", "listwise"),
  native = TRUE,
  labels = NULL
)

Arguments

Value

A psychnet object whose ⁠$weights⁠ is the symmetric standardized weight matrix, with ⁠$types⁠ and ⁠$kkt⁠ (the worst nodewise residual). A binary-binary edge carries the sign of its nodewise-logistic coefficient; mgm::mgm() reports the same edge as a magnitude only (its sign is undefined for a categorical-categorical interaction), so compare such edges on abs(). Continuous columns are standardized internally, binary predictors enter the graph on their 0/1 dummy scale, and binary-response logit coefficients are converted to mgm's two-class multinomial scale before edge aggregation. With these conventions the edge magnitudes match mgm::mgm closely for gaussian-gaussian, gaussian-binary, and binary-binary edges alike; weak edges near the EBIC/threshold boundary can still differ in support because the penalty is selected on an independent base-R path.

Examples

set.seed(1)
f <- stats::rnorm(400)
g1 <- f + stats::rnorm(400); g2 <- f + stats::rnorm(400)
b1 <- (f + stats::rnorm(400) > 0) * 1L
d <- data.frame(g1 = g1, g2 = g2, b1 = b1, n = stats::rnorm(400))
mgm_fit(d)

Description

Resamples observations with replacement, re-estimates the network on each resample, and summarizes the sampling distribution of every edge weight and node centrality (mean, percentile confidence interval, and edge inclusion proportion). An edge is flagged significant when its percentile interval excludes zero. The raw per-resample draws are stored on the returned object for use by difference_test().

Usage

net_boot(
  data,
  method = "glasso",
  n_boot = 1000L,
  ci = 0.95,
  measures = c("strength", "expected_influence"),
  centrality_fn = NULL,
  predictability = FALSE,
  threshold = FALSE,
  diff_test = FALSE,
  p_adjust = "none",
  labels = NULL,
  cores = NULL,
  engine = NULL,
  ...
)

Arguments

Value

An object of class psychnet_bootstrap: tidy ⁠$edges⁠ (with a significant flag) and ⁠$centrality⁠ data frames, the observed network in ⁠$observed⁠, raw resample draws in ⁠$edge_boot⁠, ⁠$str_boot⁠, ⁠$ei_boot⁠, and the general ⁠$centrality_boot⁠ (named list, one matrix per measure). Optional ⁠$predictability⁠, ⁠$thresholded⁠, ⁠$edge_diff_p⁠, ⁠$centrality_diff_p⁠, plus ⁠$lambda_path⁠/⁠$lambda_selected⁠ when the estimator reports them.

Examples

set.seed(1)
x <- matrix(stats::rnorm(150 * 5), 150, 5) %*% chol(0.4^abs(outer(1:5, 1:5, "-")))
colnames(x) <- paste0("V", 1:5)
bs <- net_boot(x, n_boot = 50, cores = 1)
as.data.frame(bs)

Description

Node centrality

Usage

net_centralities(
  x,
  measures = c("strength", "expected_influence"),
  centrality_fn = NULL,
  ...
)

Arguments

Value

A tidy data.frame, one row per node, with a node column and one column per requested measure (strength = sum of absolute edge weights, expected_influence = sum of signed edge weights, by default).

Examples

S <- 0.4^abs(outer(1:6, 1:6, "-"))
net_centralities(ebic_glasso(cor_matrix = S, n = 250))

Description

Permutation test for whether two groups' Gaussian graphical models differ, on three invariants: global strength (M), maximum edge difference (S), and per-edge differences (E). Networks are EBIC graphical lassos (clean-room pure R). Equivalent in purpose to NetworkComparisonTest::NCT().

Usage

net_compare(
  data1,
  data2,
  iter = 1000L,
  gamma = 0.5,
  paired = FALSE,
  abs = TRUE,
  weighted = TRUE,
  p_adjust = "none"
)

Arguments

Value

An object of class psychnet_nct with ⁠$nw1⁠, ⁠$nw2⁠, and ⁠$M⁠, ⁠$S⁠, ⁠$E⁠ (each observed, perm, p_value); ⁠$E⁠ also carries edge_names, a from/to data frame aligned to the per-edge vector.

Examples

set.seed(1)
a <- matrix(stats::rnorm(150 * 5), 150, 5)
b <- matrix(stats::rnorm(150 * 5), 150, 5)
colnames(a) <- colnames(b) <- paste0("V", 1:5)
fit <- net_compare(a, b, iter = 50)
fit

Description

A tidy, one-row-per-argument map from each reference package's estimator to its psychnet equivalent, so users migrating from qgraph::EBICglasso, qgraph::cor_auto, qgraph::ggmModSelect, IsingFit::IsingFit, or mgm::mgm can find the matching argument and see what psychnet changes by default. Cross-sectional estimators only (temporal models are out of scope).

Usage

net_crosswalk(
  reference = c("all", "EBICglasso", "cor_auto", "ggmModSelect", "IsingFit", "mgm")
)

Arguments

Value

A tidy data.frame, one row per argument, with columns reference (the pkg::fn being substituted), psychnet (the psychnet verb), ref_arg, psychnet_arg ("-" when there is no counterpart), status (identical / renamed / ⁠default differs⁠ / ⁠semantics differ⁠ / ⁠reference only⁠ / ⁠psychnet only⁠), and a short note.

Examples

net_crosswalk("EBICglasso")
net_crosswalk("IsingFit")

Description

Reports how well each node is predicted by the others in a fitted network. For Gaussian graphical models this is the closed-form variance explained (R-squared) from the precision matrix and needs no data. For the nodewise models (ising_fit(), ising_sampler(), mgm_fit()) it requires the data and reports R-squared for Gaussian nodes and classification accuracy (CC) plus normalized accuracy (nCC) for binary nodes.

Usage

net_predict(x, data = NULL, ...)

Arguments

Value

A tidy data.frame, one row per node, with columns node, type ("gaussian" or "binary"), metric ("R2" or "nCC"), predictability, and accuracy (classification accuracy for binary nodes, NA for Gaussian).

Examples

S <- 0.4^abs(outer(1:6, 1:6, "-"))
net_predict(ebic_glasso(cor_matrix = S, n = 250))

Description

Centrality-stability coefficient (case-dropping subset bootstrap)

Usage

net_stability(
  data,
  method = "glasso",
  measures = c("strength", "expected_influence"),
  centrality_fn = NULL,
  drop_prop = seq(0.1, 0.9, by = 0.1),
  iter = 100L,
  threshold = 0.7,
  certainty = 0.95,
  labels = NULL,
  ...
)

Arguments

Value

An object of class psychnet_stability with ⁠$cs⁠ (CS-coefficient per measure) and a tidy ⁠$table⁠ of mean correlations by drop proportion.

Examples

set.seed(1)
x <- matrix(stats::rnorm(200 * 5), 200, 5) %*% chol(0.4^abs(outer(1:5, 1:5, "-")))
colnames(x) <- paste0("V", 1:5)
cs <- net_stability(x, drop_prop = c(0.3, 0.5, 0.7), iter = 20)
cs$cs

Description

A thin companion to net_predict() that returns predictability as a plain numeric vector in node order, clamped to ⁠[0, 1]⁠ – the form cograph::splot() expects for pie_values (the predictability ring drawn around each node). Use net_predict() when you want the full tidy table.

Usage

node_predictability(x, data = NULL)

Arguments

Value

A named numeric vector, one value per node (node order), each in ⁠[0, 1]⁠.

Examples

S <- 0.4^abs(outer(1:6, 1:6, "-"))
node_predictability(ebic_glasso(cor_matrix = S, n = 250))

Description

Conditional (full-order) association network: each edge is the correlation between two variables with all others partialled out, obtained from the inverse correlation matrix. Equivalent to bootnet's "pcor" default.

Usage

pcor_network(
  data = NULL,
  cor_matrix = NULL,
  n = NULL,
  cor_method = c("pearson", "spearman", "kendall", "auto"),
  threshold = 0,
  alpha = NULL,
  adjust = "none",
  na_method = c("pairwise", "listwise"),
  labels = NULL
)

Arguments

Value

A psychnet object whose ⁠$weights⁠ is the thresholded partial-correlation matrix, with ⁠$precision⁠, ⁠$cor_matrix⁠ (and ⁠$p_values⁠ when alpha is used).

Examples

x <- matrix(stats::rnorm(200 * 4), 200, 4)
pcor_network(x)
pcor_network(x, alpha = 0.05, adjust = "holm")

Description

Print a psychnet network

Usage

## S3 method for class 'psychnet'
print(x, ...)

Arguments

Value

x, invisibly.

Description

Print a network bootstrap

Usage

## S3 method for class 'psychnet_bootstrap'
print(x, ...)

Arguments

Value

x, invisibly.

Description

Print a moderated MGM fit

Usage

## S3 method for class 'psychnet_moderated'
print(x, ...)

Arguments

Value

x, invisibly.

Description

Print a Network Comparison Test

Usage

## S3 method for class 'psychnet_nct'
print(x, ...)

Arguments

Value

x, invisibly.

Description

Print a centrality-stability result

Usage

## S3 method for class 'psychnet_stability'
print(x, ...)

Arguments

Value

x, invisibly.

Description

The package's main entry point: routes to the requested estimator and returns a common psychnet object, so callers can swap estimators without rewiring downstream code.

Usage

psychnet(
  data,
  method = c("glasso", "cor", "pcor", "ising", "mgm", "huge", "ggm", "tmfg", "logo",
    "relimp", "ising_sampler"),
  threshold = 0,
  gamma = NULL,
  labels = NULL,
  ...
)

Arguments

Details

method speaks the package's own short vocabulary – "glasso", "ggm", "tmfg", "logo", "relimp", "ising", "ising_sampler", "huge", "mgm", "cor", "pcor". For interoperability it also accepts the qgraph/bootnet spellings ("EBICglasso", "ggmModSelect", "TMFG", "LoGo", "IsingFit", "IsingSampler"), which resolve to the same estimators. Whichever you pass in, the stored ⁠$method⁠ is the short name.

Value

A psychnet object.

Examples

x <- matrix(stats::rnorm(200 * 5), 200, 5)
psychnet(x, method = "glasso")
psychnet(x, method = "pcor", cor_method = "spearman")
psychnet(x, method = "EBICglasso")  # qgraph alias, same result

Description

Builds a directed network in which the edge predictor -> outcome is the predictor's LMG (Shapley) share of the outcome node's regression R-squared. Equivalent in purpose to relaimpo::calc.relimp(type = "lmg") applied nodewise / bootnet's "relimp" default, pure base R and self-certified via lmg_certificate().

Usage

relimp_network(
  data = NULL,
  cor_matrix = NULL,
  cor_method = c("pearson", "spearman", "kendall", "auto"),
  max_nodes = 21L,
  na_method = c("pairwise", "listwise"),
  labels = NULL
)

Arguments

Value

A psychnet object whose ⁠$weights⁠ is the directed importance matrix (weights[k, j] = importance of k for outcome j), with ⁠$r2⁠ (per-node full-model R-squared), ⁠$cor_matrix⁠, and ⁠$kkt⁠ (the decomposition residual).

Examples

S <- 0.4^abs(outer(1:5, 1:5, "-"))
relimp_network(cor_matrix = S)

Description

Five data sets of Motivated Strategies for Learning Questionnaire (MSLQ) construct scores, each a random sample of 300 cases generated by one large language model in the study "Delving into the psychology of Machines: Exploring the structure of self-regulated learning via LLM-generated survey responses" (Computers in Human Behavior, 2025). One data set per model: SRL_GPT, SRL_Gemini, SRL_Claude, SRL_Mistral, and SRL_LLaMa.

Usage

SRL_GPT

SRL_Gemini

SRL_Claude

SRL_Mistral

SRL_LLaMa

Format

Each is a data frame with 300 rows and 5 variables – the MSLQ construct scores (each the mean of that construct's Likert items, range 1–7):

CSU

cognitive strategy use

IV

intrinsic value

SE

self-efficacy

SR

self-regulation

TA

test anxiety

Details

Only the five models whose responses carry estimable structure are included. Two other generators from the original study (ChatGPT and LeChat) produced near-independent items (mean absolute inter-item correlation $\approx 0.03$), so their network is the empty graph; they are omitted. The five retained models span the spectrum the paper describes, from realistically structured (SRL_GPT) to strongly "over-coherent" (SRL_Gemini, SRL_Claude).

Source

Saqr, M. (2025). Delving into the psychology of Machines: Exploring the structure of self-regulated learning via LLM-generated survey responses. Computers in Human Behavior, 173, 108769. doi:10.1016/j.chb.2025.108769

Examples

# partial-correlation network of the five constructs for one model
net <- ebic_glasso(SRL_GPT)
net
net_centralities(net)

Description

Summarize a psychnet network

Usage

## S3 method for class 'psychnet'
summary(object, ...)

Arguments

Value

The tidy edge list (invisibly); prints a summary as a side effect.

Description

A TMFG has no convex objective, so its correctness is certified structurally: a valid TMFG on p >= 3 nodes has exactly 3(p - 2) edges, is connected, and is chordal (every cycle of length >= 4 has a chord). Returns a non-negative score that is 0 for a valid TMFG.

Usage

tmfg_certificate(x)

Arguments

Value

Scalar; 0 certifies a valid TMFG (correct edge count, connected, chordal), otherwise a positive integer counting the violated invariants.

Examples

set.seed(1)
x <- matrix(stats::rnorm(200 * 6), 200, 6)
tmfg_certificate(tmfg_network(x))

Description

Builds a sparse, planar, chordal association network by greedily retaining the 3(p - 2) most informative edges (Massara et al. 2016). Equivalent in purpose to NetworkToolbox::TMFG() / bootnet's "TMFG" default, pure base R; correctness is certified structurally by tmfg_certificate().

Usage

tmfg_network(
  data = NULL,
  cor_matrix = NULL,
  cor_method = c("pearson", "spearman", "kendall", "auto"),
  na_method = c("pairwise", "listwise"),
  labels = NULL
)

Arguments

Value

A psychnet object whose ⁠$weights⁠ is the filtered (signed) correlation matrix on the retained edges, with ⁠$adjacency⁠, ⁠$cliques⁠, ⁠$separators⁠ (the chordal decomposition used by logo_network()), and ⁠$cor_matrix⁠.

Examples

set.seed(1)
x <- matrix(stats::rnorm(200 * 6), 200, 6)
tmfg_network(x)