HOC as a moderator

[LPhillips]

I want to test the interaction effect of a HOC as a moderator variable. If I treat the HOC (using two stage analysis) as a continuous moderator, I find no significant interaction effect. However, if I treat the HOC as a categorical moderator (by splitting the latent variable score into two groups) I do. Can anyone shed any light? is it appropriate to use the latent variable score of a HOC as a basis for a categorical moderator analysis?

Interestingly I do find that the HOC has a partial mediation effect but I am not convinced that this makes sense theoretically.

[cringle]

I would be rather careful when playing around with higher order constructs (HOC). Not everything that is technically feasible makes sense. Theory and concepts always come first. A HOC is a multidimensional construct how would its use as a moderator or mediator makes sense or is beneficial? Which of these construct's facets account for the moderation or mediation? I suggest keeping things as simple as possible. Why don't you skip the HOC (which often do not make much sense content wise) and only focus on the lower order constructs in your model?

Best regards
Christian