Closed TDJorgensen closed 5 months ago
Oh, and the first 5 commits in this PR address some other issues we discussed (e.g., new lavTestLRT()
options).
I just updated the documentation, too.
This works really well. Tests did not reveal any issues. Thanks a lot for this PR. This was a lot of work.
lav_fit_measures()
also checks fitted model for@external$h1.model
, as doeslav_object_summary()
. Here is an example to demonstrate:Tests with unadjusted df=5 are still available from
lavTest(fit0, output = "text")
Contrary to what I speculated when we discussed this in person, I no longer expect a problem with RMRs. The issue here is that there are structural constraints in the unrestricted model's implied moments, which should be reflected by df > 0. But the unrestricted moments are the same, so residuals will be the same. Their _SE_s and CIs in
lavResiduals()
might be improved, if the customh1.model
parameters are estimated more efficiently. But that's for future research.