Bifactor Models in Mplus

[@hey_its_delilah17]

Thank you! You just saved my dissertation 🧠

[@stevemota]

In a model such as the one shown at 14:18 (Task on G, Passion, Perseverance), is it possible/appropriate to model interactions between the specific factors and the general factor to look for moderation?

[@LlewellynVanZyl]

No. The whole point of a bifactor model is to isolate the specific factors from the general factor. If you create interactions you're basically making compound interactions with the same factor. So no

[@lawrencekou7923]

why don't I get the model fit information in the output result?

[@abriendosenderos]

Amazing!

[@LlewellynVanZyl]

Thanks for the feedback! :)

[@jsswan1]

Hi! I was just running a CFA today using various types of models (1 factor, 2 factor, bifactor, etc.) and I noticed that the asterisk that you seem to indicate is needed to indicate a freely estimated loading, doesn't seem to be necessary if the factor variance attached to that loading has been constrained to 1. In other words, with my factor variance constrained at 1, I still got freely estimated loadings whether asterisks were there or not. No one seems to mention this when I look around in books or online. Thoughts? Smart mPlus I guess! Otherwise, thanks for your content!

[@LlewellynVanZyl]

In my example we only have two specific factors and your model won't converge because it's not parsimonious. So then you follow the standard procedure. First remove factor constraints, then constrain variances etc. I explained this in a previous video. Same process as with any underidentified model that doesn't converge. It's an iterative process of the three steps until your model converges. 1) Paths constrained to be equal 2) Paths freely estimated and factor variance constrained to 1 3) Paths constrained to be equal and factor variances set to 1.

[@jsswan1]

@@LlewellynVanZyl You are saying my model won't converge or yours won't? Mine did converge and results were fine with same results either way (with or without asterisks). Oh, and apologies for not clarifying, I am using a model I created - not using your example models. Sorry for crashing your channel with a random question. I havent been involved in your course or other videos.

[@usamasyed2143]

Sir, Suppose we're comparing only three models (Model 1, 2 and 3 NOT 4 and 5 ). Model 2 and 3 are exactly the same. In such a scenario, which model should we retain?