Dear Min Lee,

you have some misconceptions about what a common factor model is and, in particular, its extension to a higher-order level. A common factor is an assumed underlying common cause of a bunch of dependent variables which act as its indicators (in the same way as shadows indicate / reflect an underlying light source). Hence, your second-order factor is not a composite of the primary factors but represents your assumption that it affects these.

See here some associated threads (and also the contained references).

https://www.researchgate.net/post/Scales_Subscales_in_Confirmatory_Factor_Analysis-how_do_you_analyse_them#view=5fd0852ede0d8552fe4a3716

https://www.researchgate.net/post/Whats_the_criterion_of_the_Discriminant_Validity_between_dimensions_of_second-order_variable#view=5fa535e29a19f137757dc34c

If your model fits well, you are only testing the validity of the measurement model involving the indicator-primary factor structure. The reason is that your second-order factor has only three primary factors as reflections and, thus, is just-identified/saturated. You need at least 4 primary factors OR other external variables which substantially covary with your second-order factor (which should, by the way, be the case for any factor model).

Good luck with your work,

--Holger

Holger Steinmetz Thank you very much for your insight,

I will look more in to it.

In terms of your question about whether "Can i say that latent variable 'Sk' is able to explain .77 of CQ?" I think you have the direction of causation backwards. Instead, you should say that CQ causes (.77 x .77 or about 60%) of the variance in SK. Based on the three coefficients leading from CQ to Sk, Kn, Co, it appears that CQ and Kn are so highly related that they are almost the same thing.