Firstly, I am not a statistician so I am not even certain whether all of this make any sense. However, I deal with two second-order latent constracts within social psychology field. One of them consists of two first-order sub-dimensions, and the second one has four first-order sub-dimensions. Does it make sense to incorporate these constructs into one SEM model and then discuss causality? I am fully aware of that the causality issue is a completely different problem when analyzing statistical data, so please focus on the model itself. Thank you for any comments :-)
Hi, Lukasz, it is theoritically applicable. But when you test the goodness- of -fit, or convergent validity indices of the model, It is normally not good, and the reviewers may indicate this problem. This is why many authors just discuss specified relationships between first- order factors. Hope it is useful to you.
Second-order factors deal with the measurement model in terms of idenfifying what factors underlie mresures and first-order factors. Causal links between any constructs (whether second-order latent constructs or first-order constructs) deal with the structual model. it is methodologically appropriate. So far as you have substantive ratonales and support for the causality, it is theoretically appropriate as well..
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling