Post by ì´ì±íHello, everyone.
I have a question related with structural equation modeling.
When I run sem in amos program, I get the following error message.
How do I fix it? The error message: "The model is probably
unidentified. In order to achieve identifiability, it will probably be
necessary to impose 4 additional constraints."
Note: I have never used AMOS or done SEM. My experience
is as a skeptical reader a few times, who once looked at a
few programs.
The "probably" suggests to me that an iterative process
failed to converge. If so, it is conceivable that it would
converge with a higher limit placed on iterations. Was that
part of the error message?
However, the guess at "probably" for "4 additional constraints"
implies that you likely need more constraints. Why would it
suggest a specific number more than one?
Constraints: Simplify. Set terms (means or SDs) to be zero
(or some fixed value) or equal to each other. Or specify
patterns that reduce degrees of freedom.
Or: Is your N sufficient? Is then N in the analysis fully what
you expect, or was it reduced by Missing?
* * below, from a post by me in January, 2023 * *
Borrowing from a post by Bruce Weaver,
Try posting to the SPSSX-L mailing list that is maintained by the
University of Georgia. It is far more active than this newsgroup.
You can join and post via this website:
https://listserv.uga.edu/scripts/wa-UGA.exe?INDEX
I hope this helps.
(In recent weeks, that group has not had much activity, either,
but there will be more people reading there.)
* *
If you post there, please incorporate some mention of
N and show the syntax used, and any more info that came
with the Error message.
--
Rich Ulrich