autoSEM (version 0.1.0)

autoSEM: This function houses a number of different heuristic optimization algorithms for specification search.

Description

This function houses a number of different heuristic optimization algorithms for specification search.

Usage

autoSEM(method = "GA", data = NULL, nfac = NULL, varList = NULL, criterion = "BIC", minInd = 3, niter = "default", parallel = "no", missing = "listwise", CV = "boot", R = 100, min.improve = niter, seed = NULL, std.lv = TRUE, ...)

Arguments

method

which optimization algorithm to use. Currently, it is only recommended to use "GA" for the genetic algorithm from the GA package, "aco", an implementation of the ant colony algorithm by Ross Jacobucci, and "tabu", an implementation of the Tabu search procedure by Ross Jacobucci. The latter two algorithms are based on the book chapter by Marcoulides & Leite, 2013.

data

a required dataset to search with.

nfac

the number of factors to test.

varList

list containing the names of the variables to use from the dataset.

criterion

The fit index to use as a criterion for choosing the best model. Current options are "NCP", "RMSEA","AIC", "BIC", and "BIC2", which is the sample size adjusted BIC.

minInd

The minimum number of indicators per factor.

niter

The maximum number of iterations to use. "default" changes the number of iterations based on the algorithm used.

parallel

Whether to use the snowfall package for parallelization. Note that this is only applicable for the GA package at this time.

missing

Argument to be passed to cfa() as to what to do with missing values. Note: missing="fiml" can't be paired with CV=TRUE

CV

Whether to use cross-validation for choosing the best model. The default is to use fit indices without CV.It is currently recommended to either use FALSE or "boot". Note that "boot" will take significantly longer.

R

If using bootstrap, how many samples to take? Default is 100

min.improve

Number of iterations to wait for improvement before breaking.

seed

random seed number.

std.lv

Defaults to true. So lavaan uses all variables for each factor

...

Additional arguments to pass to cfa(). An example is is setting orth=FALSE,std.lv=TRUE.

Value

fit the fit indexsolution the solution with the best fitout returned object from optimization algorithm

Examples

library(autoSEM)
myData =  HolzingerSwineford1939[,7:15]

f1.vars <- c("x1","x2","x3","x4","x5","x6","x7","x8","x9")

out = autoSEM(method="GA",data=myData,nfac=1,
             varList=list(f1.vars),CV=FALSE,
             criterion="RMSEA",minInd=3,niter=1)
summary(out)