As far as I know, the idea of "Multiple Indicators, Multiple Causes" Models was an early formulation for what we now think of as Structural Equation Models more generally. Is there something about what you are doing that does not fit easily within the general SEM approach?
First of all, thank you for your feedback. We have done a scale development study and the latest studies recommend MIMIC models to determine differential item functioning. In other words, it is preferred to test whether there is a difference in the factor structure of the scale according to some covariates. For example, do the items function differently according to gender, is measurement invariance ensured? Our goal is to answer these questions.
Should you have any further suggestions, I would be grateful to hear them. I thank you for your time and attention.
What you are describing are interaction effects, also known as moderator variables. If you have a strong theoretical or empirical reason for believing that the measurement model might vary by some characteristic, then by all means test that hypothesis. Otherwise, so are so many possible sources of difference that you cannot possibly check them all.
We don't have such a hypothesis, but some journals may request measurement invariance. I was confused about this, but your answer has enlightened me quite a bit now.
- Badges
- Science topic