Issues with invariance testing using sem command - estat fit and cross-group constraints

[Sarah Purnell]

Hello all,

I have Stata 17.0, and I am using macOS Ventura 13.1

I am trying to conduct invariance testing of two groups on one latent variable and am having a couple issues.

First: I am trying to get the RMSEA, CFI, and TLI for the configural model but the estat fit command is giving me a "estat fit not valid" error message. The syntax for my configural model is working and my model runs correctly, but this command in particular will not work. To note, the data violated the assumption of normality and so I am using the vce(sbentler) estimation but in the stata manual this should not be an issue.

Below is my syntax for the configural model:

sem (Classism -> classism1-classism13) ///
(classism1 <- Classism _cons@0), ///
group(college) ///
variance(Classism@1) mean(Classism@0) vce(sbentler) ///
ginvariant(none)

Second: When running the metric model, I get the error message "vce(sbentler) does not currently support cross-group constraints." I am unsure as to why this is not supported, is this a stata issue or an issue with the actual math that will not allow me to do this?

Below is my syntax for the metric model:

sem (Classism -> classism1-classism13) ///
(classism1 <- Classism _cons@0), ///
group(college) mean(Classism) vce(sbentler) ///
ginvariant(mcoef)

Thank you in advance!

[Sarah Purnell]

After contacting Stata technical support, I received the following response:

"After fitting model using -sem- command, we should use the command
-estat gof, stats(all)- to request the reporting of goodness of fit
statistics like RMSEA, CFI and TLI, etc.

Currently vce(sbentler) is not allowed when there are equality
constraints across the group. So, -ginvariant(none)- is the only one
that will not error out."