Skip to contents

Bootstrap Exploratory Graph Analysis

Source: R/bootEGA.R
bootEGA.Rd

bootEGA Estimates the number of dimensions of iter bootstraps using the empirical zero-order correlation matrix ("parametric") or "resampling" from the empirical dataset (non-parametric). bootEGA estimates a typical median network structure, which is formed by the median or mean pairwise (partial) correlations over the iter bootstraps (see Details for information about the typical median network structure).

Usage

bootEGA(
  data,
  n = NULL,
  corr = c("auto", "cor_auto", "cosine", "pearson", "spearman"),
  na.data = c("pairwise", "listwise"),
  model = c("BGGM", "glasso", "TMFG"),
  algorithm = c("leiden", "louvain", "walktrap"),
  uni.method = c("expand", "LE", "louvain"),
  iter = 500,
  type = c("parametric", "resampling"),
  ncores,
  EGA.type = c("EGA", "EGA.fit", "hierEGA", "riEGA"),
  plot.itemStability = TRUE,
  typicalStructure = FALSE,
  plot.typicalStructure = FALSE,
  seed = NULL,
  verbose = TRUE,
  ...
)

Arguments

data

Matrix or data frame. Should consist only of variables to be used in the analysis

n

Numeric (length = 1). Sample size if data provided is a correlation matrix

corr

Character (length = 1). Method to compute correlations. Defaults to "auto". Available options:

  • "auto" — Automatically computes appropriate correlations for the data using Pearson's for continuous, polychoric for ordinal, tetrachoric for binary, and polyserial/biserial for ordinal/binary with continuous. To change the number of categories that are considered ordinal, use ordinal.categories (see polychoric.matrix for more details)

  • "cor_auto" — Uses cor_auto to compute correlations. Arguments can be passed along to the function

  • "cosine" — Uses cosine to compute cosine similarity

  • "pearson" — Pearson's correlation is computed for all variables regardless of categories

  • "spearman" — Spearman's rank-order correlation is computed for all variables regardless of categories

For other similarity measures, compute them first and input them into data with the sample size (n)

na.data

Character (length = 1). How should missing data be handled? Defaults to "pairwise". Available options:

  • "pairwise" — Computes correlation for all available cases between two variables

  • "listwise" — Computes correlation for all complete cases in the dataset

model

Character (length = 1). Defaults to "glasso". Available options:

  • "BGGM" — Computes the Bayesian Gaussian Graphical Model. Set argument ordinal.categories to determine levels allowed for a variable to be considered ordinal. See ?BGGM::estimate for more details

  • "glasso" — Computes the GLASSO with EBIC model selection. See EBICglasso.qgraph for more details

  • "TMFG" — Computes the TMFG method. See TMFG for more details

algorithm

Character or igraph cluster_* function (length = 1). Defaults to "walktrap". Three options are listed below but all are available (see community.detection for other options):

  • "leiden" — See cluster_leiden for more details

  • "louvain" — By default, "louvain" will implement the Louvain algorithm using the consensus clustering method (see community.consensus for more information). This function will implement consensus.method = "most_common" and consensus.iter = 1000 unless specified otherwise

  • "walktrap" — See cluster_walktrap for more details

uni.method

Character (length = 1). What unidimensionality method should be used? Defaults to "louvain". Available options:

  • "expand" — Expands the correlation matrix with four variables correlated 0.50. If number of dimension returns 2 or less in check, then the data are unidimensional; otherwise, regular EGA with no matrix expansion is used. This method was used in the Golino et al.'s (2020) Psychological Methods simulation

  • "LE" — Applies the Leading Eigenvector algorithm (cluster_leading_eigen) on the empirical correlation matrix. If the number of dimensions is 1, then the Leading Eigenvector solution is used; otherwise, regular EGA is used. This method was used in the Christensen et al.'s (2023) Behavior Research Methods simulation

  • "louvain" — Applies the Louvain algorithm (cluster_louvain) on the empirical correlation matrix. If the number of dimensions is 1, then the Louvain solution is used; otherwise, regular EGA is used. This method was validated Christensen's (2022) PsyArXiv simulation. Consensus clustering can be used by specifying either "consensus.method" or "consensus.iter"

iter

Numeric (length = 1). Number of replica samples to generate from the bootstrap analysis. Defaults to 500 (recommended)

type

Character (length = 1). What type of bootstrap should be performed? Defaults to "parametric". Available options:

  • "parametric" — Generates iter new datasets from (multivariate normal random distributions) based on the original dataset using mvrnorm

  • "resampling" — Generates iter new datasets from random subsamples of the original data

ncores

Numeric (length = 1). Number of cores to use in computing results. Defaults to ceiling(parallel::detectCores() / 2) or half of your computer's processing power. Set to 1 to not use parallel computing

If you're unsure how many cores your computer has, then type: parallel::detectCores()

EGA.type

Character (length = 1). Type of EGA model to use. Defaults to "EGA" Available options:

  • "EGA" — Uses standard exploratory graph analysis

  • "EGA.fit" — Uses tefi to determine best fit of EGA

  • "hierEGA" — Uses hierarchical exploratory graph analysis

  • "riEGA" — Uses random-intercept exploratory graph analysis

Arguments for EGA.type can be added (see links for details on specific function arguments)

plot.itemStability

Boolean (length = 1). Should the plot be produced for item.replication? Defaults to TRUE

typicalStructure

Boolean (length = 1). If TRUE, returns the median ("glasso" or "BGGM") or mean ("TMFG") network structure and estimates its dimensions (see Details for more information). Defaults to FALSE

plot.typicalStructure

Boolean (length = 1). If TRUE, returns a plot of the typical network structure. Defaults to FALSE

seed

Numeric (length = 1). Defaults to NULL or random results. Set for reproducible results. See Reproducibility and PRNG for more details on random number generation in EGAnet

verbose

Boolean (length = 1). Should progress be displayed? Defaults to TRUE. Set to FALSE to not display progress

...

Additional arguments that can be passed on to auto.correlate, network.estimation, community.detection, community.consensus, EGA, EGA.fit, hierEGA, and riEGA

Value

Returns a list containing:

iter

Number of replica samples in bootstrap

bootGraphs

A list containing the networks of each replica sample

boot.wc

A matrix of membership assignments for each replica network with variables down the columns and replicas across the rows

boot.ndim

Number of dimensions identified in each replica sample

summary.table

A data frame containing number of replica samples, median, standard deviation, standard error, 95% confidence intervals, and quantiles (lower = 2.5% and upper = 97.5%)

frequency

A data frame containing the proportion of times the number of dimensions was identified (e.g., .85 of 1,000 = 850 times that specific number of dimensions was found)

TEFI

tefi value for each replica sample

type

Type of bootstrap used

EGA

Output of the empirical EGA results (output will vary based on EGA.type)

EGA.type

Type of *EGA function used

typicalGraph

A list containing:

  • graph — Network matrix of the median network structure

  • typical.dim.variables — An ordered matrix of item allocation

  • wc — Membership assignments of the median network

plot.typical.ega

Plot output if plot.typicalStructure = TRUE

Details

The typical network structure is derived from the median (or mean) value of each pairwise relationship. These values tend to reflect the "typical" value taken by an edge across the bootstrap networks. Afterward, the same community detection algorithm is applied to the typical network as the bootstrap networks.

Because the community detection algorithm is applied to the typical network structure, there is a possibility that the community algorithm determines a different number of dimensions than the median number derived from the bootstraps. The typical network structure (and number of dimensions) may not match the empirical EGA number of dimensions or the median number of dimensions from the bootstrap. This result is known and not a bug.

References

Original implementation of bootEGA
Christensen, A. P., & Golino, H. (2021). Estimating the stability of the number of factors via Bootstrap Exploratory Graph Analysis: A tutorial. Psych, 3(3), 479-500.

See also

itemStability to estimate the stability of the variables in the empirical dimensions and dimensionStability to estimate the stability of the dimensions (structural consistency)

Author

Hudson Golino <hfg9s at virginia.edu> and Alexander P. Christensen <alexpaulchristensen@gmail.com>

Examples

# Load data
wmt <- wmt2[,7:24]

if (FALSE) { # \dontrun{
# Standard EGA parametric example
boot.wmt <- bootEGA(
  data = wmt, iter = 500,
  type = "parametric", ncores = 2
)

# Standard resampling example
boot.wmt <- bootEGA(
  data = wmt, iter = 500,
  type = "resampling", ncores = 2
)

# Example using {igraph} `cluster_*` function
boot.wmt.spinglass <- bootEGA(
  data = wmt, iter = 500,
  algorithm = igraph::cluster_spinglass,
  # use any function from {igraph}
  type = "parametric", ncores = 2
)

# EGA fit example
boot.wmt.fit <- bootEGA(
  data = wmt, iter = 500,
  EGA.type = "EGA.fit",
  type = "parametric", ncores = 2
)

# Hierarchical EGA example
boot.wmt.hier <- bootEGA(
  data = wmt, iter = 500,
  EGA.type = "hierEGA",
  type = "parametric", ncores = 2
)

# Random-intercept EGA example
boot.wmt.ri <- bootEGA(
  data = wmt, iter = 500,
  EGA.type = "riEGA",
  type = "parametric", ncores = 2
)} # }