lavaan 0.6-4 has been released

[Yves Rosseel]

Dear lavaan users,

Lavaan 0.6-4 has been released and is now available on CRAN. To install, simply type in R:

install.packages("lavaan")

and restart your R session.

The release notes can be found here: http://lavaan.ugent.be/history/dot6.html

Yves.

[Jarrett Byrnes]

This is awesome. Can you elaborate to the listserv what is Structural After Measurement?

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/83890bc4-146b-4295-8428-821ae77e31ec%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Yves Rosseel]

On 7/8/19 1:59 PM, Jarrett Byrnes wrote:
> This is awesome. Can you elaborate to the listserv what is Structural
> After Measurement?

It is only relevant if you have a model with multiple (measured) latent
variables, and the focus is on the structural part: the regressions
among those latent variables.

In a nutshell, it is an implementation of an old idea (Burt, 1973, 1976)
that we should first estimate the measurement part(s) of the model, and
only then (keeping the measurement parameters fixed) estimate the
structural part of the model. This is in contrast with the standard
approach, where all parameters are estimated jointly (system-wide).

Unfortunately, there is no paper yet: I am currently writing this up.
However, special cases of the SAM approach are known as 'factor score
regression with Croon's correction'. Below are a few references.

Yves.


Burt, R. S. (1973). Confirmatory factor-analytic structures and the
theory construction process. Sociological Methods & Research, 2(2), 131-190.

Burt, R. S. (1976). Interpretational confounding of unobserved variables
in structural equation models. Sociological methods & research, 5(1), 3-52.

Devlieger, I., Mayer, A., & Rosseel, Y. (2016). Hypothesis testing using
factor score regression: A comparison of four methods. Educational and
psychological measurement, 76(5), 741-770.

Devlieger, I., & Rosseel, Y. (2017). Factor score path analysis.
Methodology, 13, 31-38.

Devlieger, I., Talloen, W., & Rosseel, Y. (2019). New Developments in
Factor Score Regression: Fit Indices and a Model Comparison Test.
Educational and Psychological Measurement, 0013164419844552.

Takane, Y., & Hwang, H. (2018). Comparisons among several consistent
estimators of structural equation models. Behaviormetrika, 45(1), 157-188.

[Jarrett Byrnes]

This is fantastic, as it might be highly useful for pircewise approaches. I’ve been noticing Thant when we use factor scores of single factor fits, the results often differ from lavaan, but I can make lavaan also produce different results if I fit measurement versus whole models as well. Fantastic - could really provide some parity!

Out of curiosity, as we think about this in the piecewise front, our current take would be to fit an individual factor for each lv. The. Use factor scores for the structural portion of the model. This makes sense when we have many nonlinear or Nonnormal relationships. Here it sounds like one should estimate the entirety of the measurement model with LVs freely covarying and then the structural part. I’m curious about literature that looks at the contrast of the two approaches and if one is a better/more valid idea.

> --
> You received this message because you are subscribed to the Google Groups "lavaan" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
> To post to this group, send email to lav...@googlegroups.com.
> Visit this group at https://groups.google.com/group/lavaan.

> To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/d0a0bd6c-8ba9-8c38-7bd1-b2535a10ab0f%40gmail.com.

[Yves Rosseel]

On 7/8/19 3:49 PM, Jarrett Byrnes wrote:
> Out of curiosity, as we think about this in the piecewise front, our
> current take would be to fit an individual factor for each lv. The.
> Use factor scores for the structural portion of the model.

Note that factor scores still contain measurement error, and using them
in a regression (without any correction) will lead to serious bias. That
is why the factor score regression approach needs to be complemented
with an additional technique, like Croon's correction, to remove this bias.

But Croon's method works on the variance/covariance matrix of the factor
scores, not the factor scores themselves. This makes it non-trivial to
extend to nonlinear relationships. The same is true for the SAM approach.

Yves

--
Yves Rosseel -- http://www.da.ugent.be
Department of Data Analysis, Ghent University
http://lavaan.org