In AnthonyRaborn/mecfa: Metaheuristic Exploratory/Confirmatory Factor Analysis

r start.time <- Sys.time()

mecfa: Metaheuristic Exploratory/Confirmatory Factor Analysis

An R package which utilizes metaheuristic algorithms to simultaneously find the best-fitting model for various factor structures and select which factor structure has the highest fit. Currently, simulated annealing is the only metaheuristic algorithm in use, but ant colony optimization may be incorporated in the future.

Installation

To install this package, you must have devtools installed! It is only found on GitHub at the moment.

require(devtools)
devtools::install_github("AnthonyRaborn/mecfa") # automatically installs stable branch

Usage

The goal of mecfa is to find the best-fitting confirmatory factor analysis from a range of possible latent factors. (If you are unsure of what this means, here is an article which talks about the basics). It fits these CFA models using the lavaan software.

To start, you need a dataset, some initial models, and the mecfa package.

System Information

sessionInfo()

Data

library(mecfa)

# sample data to use
testData <-
  lavaan::HolzingerSwineford1939

The lavaan Holzinger-Swineford dataset is a 9-item test of mental ability for 7th and 8th-grade children. It also includes other variables which may be of interest for structural equation modeling, but we will ignore these for now.

The test items are labeled x1 thru x9, so we can use these as our item names when we create the different initial models. While the generally-accepted model has 3 factors, we will investigate 2, 3, and 4 factors for this example.

Initial Models

factorsNames <-
  c("visual", "textual", "speed", "accuracy")
itemNames <-
  paste0("x", 1:9)
initialModels <-
  createInitialModels(
    maxFactors = factorsNames,
    items = itemNames)

cat(
  paste(
    initialModels,
    collapse = "\n\n"
  )
)

Function Specifications

maxSteps <-
  750
fitStatistic <-
  "rmsea"
maximize <-
  FALSE
lavaan.model.specs <-
  list(
    model.type = "cfa", 
    auto.var = TRUE, 
    estimator = "default",
    ordered = NULL,
    int.ov.free = TRUE, 
    int.lv.free = FALSE, 
    std.lv = TRUE, 
    auto.fix.first = FALSE, 
    auto.fix.single = TRUE, 
    auto.cov.lv.x = TRUE, 
    auto.th = TRUE, 
    auto.delta = TRUE, 
    auto.cov.y = TRUE,
    auto.efa = TRUE,
    missing = "listwise"
    )
maxChanges <-
  2
temperatureFunction <-
  "quadratic"

Execute Function

example <-
  exploratorySA(
    initialModels = initialModels,
    originalData = testData,
    maxSteps = maxSteps,
    fitStatistic = fitStatistic,
    maximize = maximize,
    lavaan.model.specs = lavaan.model.specs,
    temperature = temperatureFunction
  )

Investigate Results

# show/print method
example

# summary method
summary(example)

# plot method
plot(example)

It took a total of r round(as.difftime(Sys.time() - start.time, units = "mins"),2) r attr(round(as.difftime(Sys.time() - start.time, units = "mins"),2), 'units') to compile this document.

AnthonyRaborn/mecfa documentation built on Dec. 17, 2021, 8:53 a.m.