[Raphael Lefranc]

Dear Erik, Richard,

Thank you for your reply. I still want the second-order model.

I will check discussions regarding the bifactor with Stata because I want to estimate also this model.

[Richard Williams]

Hi Raphael. You can estimate the 2nd order model, but remember, it is incompatible with other goals you have stated, like uncorrelated latent variables and latent variables that all have variances of 1.

I am curious why you want it. At best, it gives you a slightly more parsimonious model, saving you a few degrees of freedom if it fits ok, as opposed to just having 4 freely correlated latent variables. But, perhaps you think there are important theoretical reasons for estimating it.

We don't actually know much about your theory or what you want to do! But hopefully all this advice will be helpful to you.

[Raphael Lefranc]

Dear Richard,

I want simply to test a more parsimonious model as you said.

More precisely I am interested in a confirmatory tetrad analysis using a second-order model as well as a bifactor model.

In order to do so I want to use the macro of Bauldry and Bollen which is perhaps the only way to do do such a thing at the moment and this is why I use Stata and not Mplus or Lavaan.

https://psycnet.apa.org/record/2016-49545-012

I have just post something about that :

https://www.statalist.org/forums/for...stat-framework

Perhaps you could be interested and help me again because I am bit lost with Stata despite my readings of documentation...

Thanks again for your help and your helpful comments.