Identification of second-order CFA

[Ferdi Botha]

I am trying to estimate a second-order CFA model. The second-order factor, A, has two first-order factors, B and C. B and C each have eight items. I am aware that a second-order model with two first-order factors is not identified. However, the model is what the theory suggests. Would it be appropriate to fix both second-order loadings to 1? Or are there alternative options available?

[ben earnhart]

Your second-order model is under-identified. It's the equivalent of a two-item single-order CFA. Need at least three items at level one to identify the second order..

[wbuchanan]

You need to provide additional information. Without knowing what constraints are being applied to the model and how many parameters are freely estimated no one can tell you whether or not the model is identified. For example, you can use "phantom" latent variables to construct single indicator latent variables that are empirically identified by constraining a sufficient number of parameters.

[Ferdi Botha]

The model is basically something like:
sem (A -> B@1 C@1) (B -> x1 x2 x3 x4 x5 x6 x7 x8) (C -> x9 x10 x11 x12 x13 x14 x15 x16). A, B, and C are latent variables. The constraints were added to the second-order loadings in order to identify the model. Again, not sure if this is appropriate? Or even necessary? As someone else also noted, when this is done it is exactly the same as if one did a two-item single-order CFA. Is it then not perhaps just best to do the latter?

[Nicole Carbert]

Hi there,
I ran the equation above with my data but it does not seem to converge. Any suggestions?
Thanks!
Nicole