Hello,

As you indicated that you deal with a reflective scale measure, I think the Factor Analysis will be more appropriate; because, we know under this scale we expected to have high inter-correlations.

Also, a SEM models (Path models) will be an other issue for such problem.

Best regards, and waiting for our colleagues adding to this.

Hello - can you provide more details about the construct since you refer to both formative and reflective constructs in your question. In particular, do you conceptually think of brand perception as a formative construct or a reflective construct? The way you perceive or hypothesize it will impact the structure of your SEM model and likewise the parameterization. Either way, your solution lies within SEM approach.

In a formative construct, the indicators (4 in your case) cause the construct. Thus, in this case your indicators will ‘cause’ (so to speak) your latent construct of brand perception, and which inturn will affect other endogenous variable(s) (not sure if you’ve other endogeneous variable, or is it just this ‘formative’ construct?)

On the other hand, reflective constructs are more like the conventional latent variables approach, where the indicators are caused BY the latent variable (which could be perception of the brand if you believe it is a reflective construct). In this case, assuming there is no other observed endogeneous variable, a measurement model would suffice. Of course, you want to include an indicator for the two groups (treatment/control). At a next step, given the significant amount of heterogeneity in consumers’ perceptions or purchase decisions, I would strongly recommend to go for a heterogeneity-based variant of the conventional SEM, or a mixed SEM at minimum.

Best, wali