qgraph (version 1.5)

qgraph.cfa: CFA using Structural Equation Modelling

Description

This function performs a simple confirmatory factor analysis using sem (Fox, 2010) or lavaan (Rosseel, 2011).

Usage

qgraph.cfa(S, N, groups=NULL, ..., pkg = "sem", labels=NULL, fun = qgraph, opts = list())

Arguments

S

A covariance matrix

N

The number of observations

groups

The groups list, see qgraph. This must be a list in which each element is a factor indicating which variables load on the same factor

…

Arguments passed to 'fun'

pkg

A string indicating which package should be used for estimating the model. Currently "lavaan" and "sem" (default) are supported

labels

A vector indicating the label of each variable

fun

A function to which the results are send. Defaults to qgraph, but can be any function that can handle the output (e.g. qgraph.lavaan, qgraph.sem, summary, print). If pkg="sem" then this can also be qgraph.loadings.

opts

A list containing arguments that are sent to either sem or cfa

Value

A "sem" object, see sem

Details

This function can be used to perform a simple confirmatory factor analasys using regular qgraph input. The function computes a model and then sends it to sem (sem; Fox, 2010) or cfa (lavaan; Rosseel, 2011). based on the package used either a "sem" object or a "lavaan" object is returned that can be used for manual inspection or to sent to qgraph.sem or qgraph.lavaan.

The model that is estimated is a first order factor model in which each variable loads on one factor and the factors are correlated. This model is specified with the `groups` argument. This must be a list in which each element represents a factor. Each element of the list must be a vector indicating which variables load on the same factor. The model is identified by fixing the first loading of each factor to 1, which should be an identifying restriction if there are at least 4 variables per factor.

The function also sends its results to another function for visualization. If this is qgraph, the default, then a visualization of the standardized coeficients is plotted.

Currently the sem package is better supported in qgraph, but this will change in a future version. Using the lavaan package can greatly reduce computation time.

References

Sacha Epskamp, Angelique O. J. Cramer, Lourens J. Waldorp, Verena D. Schmittmann, Denny Borsboom (2012). qgraph: Network Visualizations of Relationships in Psychometric Data. Journal of Statistical Software, 48(4), 1-18. URL http://www.jstatsoft.org/v48/i04/.

John Fox with contributions from Adam Kramer <[e-mail]> and Michael Friendly (2010). sem: Structural Equation Models. R package version 0.9-21. http://CRAN.R-project.org/package=sem

See Also

qgraph qgraph.sem qgraph.lavaan qgraph.loadings qgraph.semModel sem

Examples

# NOT RUN {
# Simulate dataset:
set.seed(2)
eta<-matrix(rnorm(200*5),ncol=5)
lam<-matrix(rnorm(50*5,0,0.15),50,5)
lam[apply(diag(5)==1,1,rep,each=10)]<-rnorm(50,0.7,0.3)
th<-matrix(rnorm(200*50),ncol=50)
Y<-eta%*%t(lam)+th

# Create groupslist
gr<-list(1:10,11:20,21:30,31:40,41:50)

# Using "lavaan" package:
res <- qgraph.cfa(cov(Y),N=200,groups=gr,pkg="lavaan",vsize.man=2,vsize.lat=10)

qgraph.lavaan(res,filename="lavaan",legend=FALSE,groups=gr,edge.label.cex=0.6)

# Using "sem" package:
res <- qgraph.cfa(cov(Y),N=200,groups=gr,pkg="sem",vsize.man=2,vsize.lat=10,fun=qgraph.loadings)

qgraph.semModel(res,edge.label.cex=0.6)

qgraph(res,edge.label.cex=0.6)

qgraph.sem(res,filename="sem",legend=FALSE,groups=gr,edge.label.cex=0.6)

### Big 5 dataset ###
data(big5)
data(big5groups)

fit <- qgraph.cfa(cov(big5),nrow(big5),big5groups,pkg="lavaan",opts=list(se="none"),
	vsize.man=1,vsize.lat=6,edge.label.cex=0.5)
print(fit)
# }