Dear Dr. Palij

Thank you very much for your answer. I have sent a message to you in this regard.

Thanks again

Dear Dr. Palij

Thank you very much . Now I clearly realized the answer to my question. I won't forget your valuable help. Based on your description, it seems that some of papers about path analysis published in journals may have some flaws because they do not consider reciprocal relationships while, reciprocal relationships almost always  exist and are significant.

If you let me, I would like to use some parts of your very interesting discussion in my paper that I've spoken about it in my message (Of course, along with Acknowledgement).

Thanks again

Dear Kayvan et all, 

When you aplly second order factor, in AMOS you make a new factor that connect some fisrt order factors, but you need to include "1" in one of the paths between this factor and one of its 1st order components. In this way these 1st order factors become having a error term associated, because they act now as dependent variabels. Nex you should have included covariates (rows in both directions) between the second order factor and all the others factors (this do not include the 1st order factors connected to the 2nd order). You should calculate CR and AVE in order to analyse realiability and validity.

Have a nice time

Helena

Dear Dr. Helena

Thank you so much

Yes true, that is my main concern. In fact, the first-order variables play two fully featured roles simultaneously, 1) they act as response variable for the second-order predictors and at the same time, 2) they play the role of predictor variable for the main response variable. 

In some papers I saw that the authors manually integrated path coefficients of first- and second-ordered predictor variables into a single graph (model). We should note that these coefficients and their corresponding error terms have been obtained via applying some separate standardized regressions.

Therefore, the main question would be something like this:

Are path coefficients and their corresponding error values obtained in separate standardized SMLR models still valid when we integrate them manually into a single model?

I think the answer is "NO" because, as you mentioned,  the first-order predictor variables play two different roles at the same time that we can't see such a role in separate regression analyses.

Therefore, I have to say that, AMOS calculates path coefficients exactly like standardized regression analysis for first- or second-order predictors separately. However, when we complicate the model, (I mean when we have both first- and second-order predictor variables in the model at the same time) the AMOS  calculates path coefficients using a more complicated model. Hence, the outcome can't be equal as compared to the separate standardized regression coefficients.

Kayvan

Dear Kayvan, 

Then we both think in the same way about AMOS .

Have a nice time

Helena