Type:	Package
Title:	Building Sets of Variables in a Probabilistic Framework
Version:	1.0.0
Description:	Create sets of variables based on a mutual information approach. In this context, a set is a collection of distinct elements (e.g., variables) that can also be treated as a single entity. Mutual information, a concept from probability theory, quantifies the dependence between two variables by expressing how much information about one variable can be gained from observing the other. Furthermore, you can analyze, and visualize these sets in order to better understand the relationships among variables.
License:	CC BY 4.0
Depends:	R (≥ 4.1.0)
Imports:	dplyr (≥ 1.1.4), igraph (≥ 2.1.2), permutes (≥ 2.8), pheatmap (≥ 1.0.13), splitTools (≥ 1.0.1)
Encoding:	UTF-8
LazyData:	true
RoxygenNote:	7.3.2
Suggests:	knitr, rmarkdown, testthat (≥ 3.0.0)
Config/testthat/edition:	3
VignetteBuilder:	knitr
URL:	https://github.com/nicolasleenaerts/setweaver
NeedsCompilation:	no
Packaged:	2026-02-01 18:17:26 UTC; u0127988
Author:	Nicolas Leenaerts [aut, cre, cph], Aaron Fisher [aut, cph]
Maintainer:	Nicolas Leenaerts <nicolas.leenaerts@kuleuven.be>
Repository:	CRAN
Date/Publication:	2026-02-04 18:00:08 UTC

ce

Description

Computes the conditional entropy H(Y \mid X) for two binary vectors 'y' (outcome) and 'x' (predictor).

Usage

ce(y, x)

Arguments

Value

A numeric scalar giving H(Y \mid X).

Examples

ce(misimdata$y,misimdata$x1)

cprob

Description

Computes the conditional probability P(Y=1 \mid X=1) for two binary vectors 'y' and 'x'. Rows with missing values in either vector are excluded.

Usage

cprob(y, x)

Arguments

Value

A numeric scalar giving the conditional probability that 'y = 1' given 'x = 1'.

Examples

cprob(misimdata$y,misimdata$x1)

cprob_inv

Description

Computes the conditional probability P(Y = 1 \mid X = 0) for two binary vectors 'y' and 'x'. Rows with missing values in either vector are excluded.

Usage

cprob_inv(y, x)

Arguments

Value

A numeric scalar giving the conditional probability that 'y = 1' given 'x = 0'.

Examples

cprob_inv(misimdata$y,misimdata$x1)

entfuns

Description

Computes a set of descriptive diagnostics for a binary outcome 'y' against one or more predictors in 'x', including marginal probability, conditional probability, absolute and proportional differences between marginal and conditional probabilities, and analogous measures based on . entropy.

Usage

entfuns(y, x)

Arguments

Details

Inputs are treated as binary (0/1 or logical). Missing values are removed pairwise for each predictor (rows with 'NA' in either the outcome or the predictor are excluded for that predictor's calculations).

Value

A data frame with one row per predictor and the following columns:

xvar

Predictor name.

yprob

Marginal probability P(Y=1) computed on complete cases for that predictor.

xprob

Marginal probability P(X=1).

cprob

Conditional probability P(Y=1 \mid X=1).

cpdif

Absolute difference P(Y=1 \mid X=1) - P(Y=1).

cpdifper

Percent difference relative to P(Y=1).

yent

Entropy H(Y).

ce

Conditional entropy H(Y \mid X).

cedif

Absolute difference H(Y) - H(Y \mid X).

cedifper

Percent difference in entropy relative to H(Y).

Examples

entfuns(misimdata$y,misimdata[,2:5])

entropy

Description

Returns marginal entropy for binary variables

Usage

entropy(x)

Arguments

Value

A numeric scalar giving the entropy of 'x'.

Examples

entropy(misimdata$x1)

find_minimal_sets

Description

Given a character vector of sets (each set encoded as variable names joined by a separator), returns the subset of sets that are minimal: no returned set is a strict superset of another. Duplicates and ordering differences are handled according to the implementation.

Usage

find_minimal_sets(str_vec, sep = "_")

Arguments

Value

A character vector containing the minimally sufficient sets (i.e., sets that are not strict supersets of any other set in 'str_vec').

Examples

pairmiresult = pairmi(misimdata[,2:6])
results_probstat <- probstat(misimdata$y,pairmiresult$expanded.data,nfolds=5)
find_minimal_sets(results_probstat$xvars[results_probstat$cprob >= 0.20])

gstat

Description

Computes the likelihood-ratio test statistic (G statistic) from the mutual information and the joint count of two variables:

G = 2 \times n \times MI,

where n is the joint sample size and MI is the mutual information.

Usage

gstat(mi, count)

Arguments

Value

A numeric scalar giving the G statistic value.

Examples

gstat(mi(misimdata$y,misimdata$x1),jtct(misimdata$y,misimdata$x1))

joint

Description

Computes the joint probability P(X = 1, Y = 1) for two binary vectors 'x' and 'y'. Rows with missing values in either vector are excluded.

Usage

joint(y, x)

Arguments

Value

A numeric scalar giving the joint probability that both 'x = 1' and 'y = 1', calculated as the joint count divided by the number of complete cases.

Examples

joint(misimdata$y,misimdata$x1)

jtct

Description

Counts the number of complete observations where both a binary outcome 'y' and a binary predictor 'x' equal 1. Missing values are excluded pairwise (rows with 'NA' in either 'x' or 'y' are ignored).

Usage

jtct(y, x)

Arguments

Value

An integer scalar giving the number of observations where 'x == 1' and 'y == 1', after excluding missing values.

Examples

cprob_inv(misimdata$y,misimdata$x1)

mi

Description

Computes the mutual information (MI) between an outcome 'y' and a predictor 'x', using the standard definition:

MI(X, Y) = H(X) + H(Y) - H(X, Y),

Usage

mi(y, x)

Arguments

Value

A numeric scalar giving the mutual information between 'x' and 'y'

Examples

mi(misimdata$y,misimdata$x1)

misimdata

Description

A data set with 10 predictors and 1 outcome that can be used to try out the functions of the setweaver package

Usage

misimdata

Format

A data frame with 2500 rows and 11 variables:

y

Outcome

x1

First binary predictor

x2

Second binary predictor

x3

Third binary predictor

x4

Fourth binary predictor

x5

Fifth binary predictor

x6

Sixth binary predictor

x7

Seventh binary predictor

x8

Eighth binary predictor

x9

Ninth binary predictor

x10

Tenth binary predictor

pairmi

Description

A function that calculates the mutual information for sets of variables, calculates the G statistic, determines the significance of the sets, and only keeps those that are significant.

Usage

pairmi(data, alpha = 0.05, MI.threshold = NULL, n_elements = 5, sep = "_")

Arguments

Value

A list with the following components:

expanded.data

A data frame containing the original variables and the columns for significant sets (e.g., pair/triplet indicators).

original.variables

Character vector of the original variable names.

sets

A data frame describing significant sets, including their members, size, MI, G statistic, p-value, and constructed name.

Examples

pairmi(misimdata[,2:6])

plot_prob

Description

Creates a network-style graph showing how a set of predictors ('x_vars') are related to an outcome ('y_var'). Relationships can be displayed either as conditional probabilities or as effects estimated by logistic regression.

Usage

plot_prob(
  data,
  y_var,
  x_vars,
  var_labels = NULL,
  prob_digits = 2,
  method = "conditional",
  title = NULL,
  vertex_color = "lightblue",
  vertex_frame_color = "darkblue",
  vertex_label_color = "black",
  edge_color = "darkgrey",
  edge_label_color = "black",
  min_arrow_width = 1,
  max_arrow_width = 10,
  node_size = 45,
  label_cex = 0.8
)

Arguments

Value

A graph object (typically an ['igraph::igraph'] object or similar) is returned and plotted. Nodes represent variables and edges represent associations. Node labels include variable names and marginal probabilities. Edge labels display either conditional probabilities or logistic regression effects.

Examples

plot_prob(misimdata,'y',colnames(misimdata[,3:6]),method='logistic')

prob

Description

Computes the marginal probability P(X = 1) for a binary vector 'x', ignoring missing values.

Usage

prob(x)

Arguments

Value

A numeric scalar giving the proportion of entries equal to 1 among the non-missing values of 'x'.

Examples

prob(c(0, 1, 1, 0, 1))

probstat

Description

Computes marginal, conditional, and information-theoretic summaries for a binary outcome 'y' against one or more predictors in 'x'. Performs either Fisher's exact test or a generalized linear mixed model (GLMM) for inference.

Usage

probstat(y, x, test = "Fisher", ri, nfolds, seed = 10101)

Arguments

Value

A data frame with one row per evaluated predictor (or pair) and the following columns:

xprob

Marginal probability of X=1.

yprob

Marginal probability of Y=1.

cprob

Conditional probability P(Y=1 \mid X=1).

cprobx

Conditional probability P(X=1 \mid Y=1).

cprobi

Inverse conditional probability P(Y=1 \mid X=0).

cpdif

Difference P(Y=1 \mid X=1) - P(Y=1).

cpdifper

Percent difference relative to P(Y=1).

xent

Entropy of X.

yent

Entropy of Y.

ce

Conditional entropy of Y \mid X.

cedif

Difference between marginal and conditional entropy of Y.

cedifper

Percent difference in entropy.

p

p-value from Fisher's exact test or the GLMM (as applicable).

Examples

pairmiresult = pairmi(misimdata[,2:6])
probstat(misimdata$y,pairmiresult$expanded.data,nfolds=5)

setmapmi

Description

Creates a set map visualization from the output of [pairmi()], showing which original variables compose the derived sets at a specified depth.

Usage

setmapmi(original_variables = NULL, sets = NULL, n_elements = NULL)

Arguments

Value

A setmap showing which original variables make up the sets at a certain depth

Examples

pairmiresult = pairmi(misimdata[,2:6])
setmapmi(pairmiresult$original.variables,pairmiresult$sets,2)

zprob

Description

Computes the z-score for testing whether the proportion (probability) of successes in 'x' differs from zero.

Usage

zprob(x)

Arguments

Value

A numeric value giving the z-score for the observed proportion.

Examples

zprob(misimdata$x1)