Can AMOS produce the data for an EFA table or can this only come from SPSS?

#1
Hi all,

Thoughts please.

Obviously, AMOS is mainly used for confirmatory factor analysis (CFA) and structural equation modelling (SEM) - but can data from AMOS be used to populate an exploratory factor analysis (EFA) table as part of a measure development process? Specifically can AMOS provide data on:

1. how items load on "all" factors *
2. the % of variance factor(s) explain

[* my hunch is that AMOS can't do 1. because it only tests how items load on the factor they are deemed to load on according to the path diagram].

Why the ask? Simply trying to present a measure development process in a journal with limited words. Want to show a final EFA table. Wonder whether that data can come from AMOS, or if it needs to come from using FACTOR in SPSS.

Thanks in advance for your thoughts.

Best,
mvdct
 

spunky

Doesn't actually exist
#2
AMOS is an SEM add-on only. SPSS by default can do EFA analyses so it would be redundant for SPSS to offer the same procedures both in their basic package and the add-on.
 

mvdct

New Member
#3
AMOS is an SEM add-on only. SPSS by default can do EFA analyses so it would be redundant for SPSS to offer the same procedures both in their basic package and the add-on.
Thanks for your reply Spunky.

I totally agree. It would be redundant for AMOS to repeat SPSS functionality.

I just wondered - in addition - if there might be other, more stats based answers, to the question.

For example, as above, an SEM/CFA model wouldn't be able to replicate an EFA table because the CFA model will only test how items load on factors according to the path diagram (not how all items load on all factors).

Or, might the setting of one item / factor relationship to 1 in a CFA path diagram preclude something from an EFA point of view.

Thanks.
Best,
mvdct.
 

spunky

Doesn't actually exist
#4
For example, as above, an SEM/CFA model wouldn't be able to replicate an EFA table because the CFA model will only test how items load on factors according to the path diagram (not how all items load on all factors).
If you set a path from all factors to all the items, all the cross-loading items, all the error covariances and all the correlations among factors will run of out degrees of freedom and your model will not be identified. And if you omit even one single path, then whatever you end up with is not a pure EFA.


Or, might the setting of one item / factor relationship to 1 in a CFA path diagram preclude something from an EFA point of view.
I'm not quite sure what you mean by this. Do you mean setting a loading to 1 to identify the scale of the latent factor?

Either way, I'm not exactly sure what you're attempting to get at but, in general, you won't be able to get from a CFA/SEM model to an EFA model because CFA/SEM models are much more restricted. The closest you can get would be going from EFA to CFA by attempting something like a Target or Procrustes type of rotation.