Hello Syed,

So, you're proposing a second-order factor model for each the 21-item variable set and the 33-item variable set? (E.g., the 21 items load on four factors, which subsequently load on a common second-order factor.)

Yes, this can be done. Of course, there should be some suitable rationale for the proposed model in each instance. As well, you can evaluate the adequacy of each model to reproduce the observed relationships in the respective data sets, via SEM (specifically, CFA).

There's no real upper limit on the number of variables that can be incorporated, except for practical limitations due to logistics of data collection and analysis.

Good luck with your work.

Hi Syed,

your wording ("squeeze into a construct") suggest that you have a kind of formative model in mind or an index or a shopping list but not --as David assumed--a factor model that would claim that all indicators are caused by one underlying latent variable. If this is true, you can do what ever you want as there is not "truth" or "validity" inherent in forming index--just practical or theoretical utility.

Best,

Holger

Let me understand your problem. Let us say that you are interested in assesing socio-economic status of certain community in certain country. The socio-economic status is one construct. The socio-economic status can be assessed with the help of so many other variables (for you it may be a construct) like educational status of family members, income status of family members, land owner ship, possession of live stock. In this each is a variable or construct. So, when the information collected on above variables, is combined, we can form the opinion about the social status of the family like low, medium and high. I think your problem can be more or less similar.

If yes, please indicate, then I can tell you the way, I have proceeded in my socioeconomic status project.

Hi Syed Ahsan Ali Zaman

You may start with measurement theory and theoretically establish your construct of interest. Also, check Chapter 1: https://www.smartpls.com/documentation/getting-started/pls-sem-book

I agree that what you are specifying is a second-order confirmatory factor analysis (CFA), and this would not require any "squeezing." In addition, if you have independent and dependent variables, then this CFA will be part of a structural equation model (SEM).

Hi all,

he does not specify any factor structured model at all. A "construct" can be anything--factor models (primary or secondary) have a very rigid and determined causal structure in which the correlation among indicators is said to result from the causal operation of an underlying common variable.

The non-specified way, he presents his problem concurs with any umbrella term-like construct.

Don't step in to the factor-trap.

Best,

Holger

Holger Steinmetz I was assuming that his "sub-constructs" were the specification of his intermediate factor structure.

David L Morgan Since he did not respond to any of your support attempts, we will never know :)

All the best,

Holger