Using a correlation matrix as input for efaUnrotate

[studerus]

It would be nice if the efaUnrotate function could also handle correlation or covariance matrices as input for the exploratory factor analysis.

[TDJorgensen]

Have you tried it? It is just a wrapper around lavaan(), so you can use the same sample.cov= and sample.nobs= arguments.

unrotated <- efaUnrotate(HolzingerSwineford1939, nf = 3,
                         varList=paste0("x", 1:9),
                         sample.cov = cov(HolzingerSwineford1939[paste0("x", 1:9)]),
                         sample.nobs = nrow(HolzingerSwineford1939))

If you want to be able to trust your SEs, you need to analyze a covariance matrix (you can still look at the standardized solution). Otherwise, you need to constrain your residual variances such that the total implied variances are 1 (because that is fixed in a correlation matrix).

[studerus]

I tried it but experienced an error if only sample.cov is supplied without data. In your example you have provided both, in which case I think lavaan will analyze the data .

[TDJorgensen]

Ah yes, I was working too quickly, didn't remember to remove data=. I just updated efaUnrotate() to accept summary statistics. You can install the development version (also of lavaan) in a fresh R session, then restart R before loading semTools:

install.packages("lavaan", repos = "http://www.da.ugent.be", type = "source")
devtools::install_github("simsem/semTools/semTools")
## restart R
library(semTools)
unrotated <- efaUnrotate(nf = 3,
                         sample.cov = cov(HolzingerSwineford1939[paste0("x", 1:9)]),
                         sample.nobs = nrow(HolzingerSwineford1939))
summary(unrotated, std = TRUE)

Let me know if it doesn't work for your own data.

FYI, lavaan provides some experimental support for EFA now: https://github.com/yrosseel/lavaan/issues/112 (see example at the end of the thread). So I expect efaUnrotate() will eventually be deprecated, but not until lavaan's functionality is well documented.

[studerus]

Thanks a lot for the quick fix! It works now.

[mrli1234]

Ah yes, I was working too quickly, didn't remember to remove data=. I just updated efaUnrotate() to accept summary statistics. You can install the development version (also of lavaan) in a fresh R session, then restart R before loading semTools:

install.packages("lavaan", repos = "http://www.da.ugent.be", type = "source")
devtools::install_github("simsem/semTools/semTools")
## restart R
library(semTools)
unrotated <- efaUnrotate(nf = 3,
                         sample.cov = cov(HolzingerSwineford1939[paste0("x", 1:9)]),
                         sample.nobs = nrow(HolzingerSwineford1939))
summary(unrotated, std = TRUE)

Let me know if it doesn't work for your own data.

FYI, lavaan provides some experimental support for EFA now: yrosseel/lavaan#112 (see example at the end of the thread). So I expect efaUnrotate() will eventually be deprecated, but not until lavaan's functionality is well documented.

It is possible to run Bifactor-ESEM using semTools? If yes, where can I find related examples? Thanks.

[TDJorgensen]

ESEM using semTools?

No, don't expect this to be supported by semTools.