No es mi especialidad, pero me parece que podrías probar con el análisis factorial confirmatorio

Thanks for your response Hugo, can you please translate it in English?

Dear Sattar,

At the first place you should consider all constructs as if they are first order ones (forget the second order ones) and perform measurement model (CFA), checking factor loadings, CR, AVE and AVE vs. squared correlations, and checking the fit of the model.

If you need to delete any indicator (observed variable) due to low factor loading, you can do this, staying at least with 3 per latent variable.

After that, when you define the apropriate measurement model, you introduce the second order variable, checking again model fit.

Last step is to build the structural model, establishing adequate path according to the theory background and the model you want to test.

Hope this help.

Anything else, feel free to call me back.

Regards,

Alcides Barrichello

Thanks Alcides. 

Hi Sattar,

Due the complexity of your model, i suggest the use of Partial Least Square measure model. Sotwares SmartPLS or Package lavaan - R Project (for example).

It is easy find many tutorials in internet.

Hope that help

Best regards

Dirceu

Thanks, Dirceu, for your suggestion.