Multigroup Analyses with Pairwise Comparisons

[Reece Akhtar]

Hi,

I am fairly comfortable in using lavaan to run multigroup SEM analyses, however I am running into a slight problem, that I was hoping someone could help me with.

I originally learned how to do SEM using AMOS, in which it is easy to test whether the strength of a parameter in one group is significantly different to the same parameter in another. This is done through using the critical-ratio function and working out a z-score.

I have searched message boards, papers and tutorials yet I am still unsure as to how I would do this in lavaan.

The example I am working with is a SEM that has two groups (males & females), where I am testing the effect of family life and personality on venture success. I am able to run the multigroup model in lavaan and produce estimates for both groups.

I have looked in the .inspect command hoping to find the right function, but have been unsuccessful.

Any ideas or suggestions are greatly appreciated.

Kind regards,

Reece

[Edward Rigdon]

Reece--

     You test an equality constraint by comparing model fit with and without the constraint.  So with two groups, to test equality for the path where A predicts B, the syntax:

B ~ c(dog,dog)*A

Imposes the equality constraint, while 

B ~ c(dog,cat)*A

Removes the constraint, leaving you with a chi square difference test with 1 degree of freedom.

     Are you asking about something else?

--Ed Rigdon

Sent from my iPad

--
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 http://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[Reece Akhtar]

Hi Ed,

Thank you for your reply, however I am not sure if I fully understand your suggestion/syntax. In your example, does dog represent the name of the group?

Kind regards,

Reece

[Edward Rigdon]

     In the suggested syntax, “dog” and “cat” are labels for parameters.  Using the same label for two parameters constrains them to equality.  With two groups, for each parameter there are two values.

     See the very nice tutorial at the lavaan project site:

http://lavaan.ugent.be/tutorial/groups.html

--Ed Rigdon

[Jorge Sinval]

Hello!

I have the same question, Rigdon solution is the most common procedure, fixing the parameters and do a chi-square test. But I think that Akhtar is looking (as I'm) for another approach, with the regression weights and critical ratios. Like in this video. Any thoughts?

[Terrence Jorgensen]

I have the same question, Rigdon solution is the most common procedure, fixing the parameters and do a chi-square test. But I think that Akhtar is looking (as I'm) for another approach, with the regression weights and critical ratios. Like in this video. Any thoughts?

Look into a Wald test, which only requires fitting the less constrained model.  Using the lavTestWald() function, you can pass a (set of) constraint(s) to test for equivalence.  This is equivalent to the chi-squared difference test, but only requires fitting one model.  Note that you still need to include labels for any parameters in the model syntax that you might want to test for equivalence.

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