delta/ theta parameterization

[bruno....@gmail.com]

Hi,

I am working on a CFA model with categorical variables. 

I used the wlsmv estimator and i wonder what the differennces between the parameterization types (delta/theta) are.

Am i right in assuming that the theta parametrization is the default setting (is there a source for that) in lavaan?

When do i use which parameterization?

Thanks in advance!!

Bruno

[kma...@aol.com]

Bruno,
The difference is described in the below article on page 489.

https://pdfs.semanticscholar.org/ecb9/90931736e4b1a2f6c4e5bada671142d7f2a9.pdf

You can also find discussion of the differences in the Mplus discussion boards and in Rex Kline's 4th edition of his SEM text.

Keith
------------------------
Keith A. Markus
John Jay College of Criminal Justice, CUNY
http://jjcweb.jjay.cuny.edu/kmarkus
Frontiers of Test Validity Theory: Measurement, Causation and Meaning.
http://www.routledge.com/books/details/9781841692203/

[bruno....@gmail.com]

Thanks a lot!

[Terrence Jorgensen]

Am i right in assuming that the theta parametrization is the default setting (is there a source for that) in lavaan?

No, delta is the default (constraining the total latent-item-response variance to 1).

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

UvA web page: http://www.uva.nl/profile/t.d.jorgensen

[bruno....@gmail.com]

Thanks a lot!

I am new to cfa and lavaan and trying to learn as much as i can about it.

I hope my questions aren't to trivial.

I also used the reliability function from the semTools packages and compared the different reliability estimates (McDonald's Omega, Raykov's Omega and Bentler's rho).

The estimates are very similar for my model. Is there a rule of thump when to use which of these estimates?

And what do you think of the omega function for the second order model (reliabilityL2 )?

Best regards.

Am Dienstag, 7. Februar 2017 20:27:57 UTC+1 schrieb bruno....@gmail.com:

[bruno....@gmail.com]

And can i use all of them for binary/ categorical data?

Am Dienstag, 7. Februar 2017 20:27:57 UTC+1 schrieb bruno....@gmail.com:

[Terrence Jorgensen]

And can i use all of them for binary/ categorical data?

Yes, but read the help page about details for which measures are returned for categorical indicators.

[Terrence Jorgensen]

The estimates are very similar for my model. Is there a rule of thump when to use which of these estimates?

And what do you think of the omega function for the second order model (reliabilityL2 )?

I'm not sure, but feel free to post these particular questions on SEMNET, which is a much wider forum for general SEM questions.

www2.gsu.edu/~mkteer/semnet.html