How to release constraints for Measurement Invariance in JASP
I was running the multigroup SEM analysis with JASP version 0.15 to examine measurement invariance.
The results showed that metric invariance is not supported (as the difference in CFI .01). I would like to release the constraint on one of the factor loadings (say loading of item 2) to test partial metric invariance.
Although there is a box of "Releasing constraints", I have no clue what information I shall insert to release the fixing of the factor loading of item 2. Please advise.
Comments
Btw I tried some stuff but couldn't figure it out either... am curious to see as well
I'll forward this issue to our SEM expert.
E.J.
Hey CSTan and patc3,
I've received your question(s), but answering will require some further consultation with some colleagues. I'll get back to you asap.
Best,
Michael
Following up on this thread. I also want to figure out partial invariance and have not figured out what goes in the release constaints box.
Michael has to concentrate on his studies for the moment but I'm going to find someone else to look at this...
E.J.
Hello everyone,
the Release constraints box should accept the same syntax as the group.partial argument in lavaan. See https://lavaan.ugent.be/tutorial/groups.html, section "Constraining groups of parameters to be equal across groups" for an example.
Best,
Simon
Simon,
I tried that, but it did not seem to work. I constrained regression coefficients across two groups, but tried to release one by including the line "PES ~ WPA_B" in the release constraints box. In the output, those regression coefficients still had the same prefix (.p3) for both groups being compared, and the results did not change from constraining all the regression coefficients. I must be missing something. Any ideas?
Did anyone find any solution for this problem? I have faced with the same problem. I have no idea how I may show to partial invariance...
I am expecting to encounter this issue a few times in the upcoming weeks, and may have to do the whole analysis directly in lavaan, but I'd love to keep using JASP as it is much easier.
Just thought I would follow-up on this again, as I seem to be encountering this issue frequently, and I'd love to figure out how to get JASP to handle it.
I'll bring this to Simon's attention again
Cheers,
E.J.
Having the same issue. Good to know that's being looked upon. I also recently noticed that JASP latest version cannot directly compare models for measurement invariance. You need to write a different code for every model, as selecting a constraint (Loadings, Intercepts, Residuals...) applies them for every one of them. I remember old versions allowed these checkboxes to be individual for each model. Hope that's also being looked upon.
Sorry for the tardy response!
It seems there is a bug, I filed an issue on GitHub so that we can keep track of it better: https://github.com/jasp-stats/jasp-issues/issues/1687