In my study, I have one independent (antecedent) leadership variable, two mediating variables at the individual level, and two moderating variables at the organizational level . The moderators influence the relationship between the mediators and the dependent variable. I also have two different dependent variables that represent alternative measurements of the same outcome. All variables except the two moderators are based on individual-level experiences, while the moderators are contextual/organizational. I want to build and test two parallel models—each with the same structure but a different dependent variable. I’m seeking guidance on how best to structure these models (e.g., diagram, SPSS, SEM, path analysis, multi-level modeling) and test them statistically, including whether I need to treat the moderators as cross-level interactions or something else.
The most effective and efficient way to run this model is probably through multilevel path analysis with random intercepts and random slopes. The cross-level interactions would then automatically be included in the model. The Mplus software allows you to specify twolevel path models using the ANALYSIS: TYPE = TWOLEVEL RANDOM option. You can use this technique with multiple outcomes and/or multiple mediators in a single model/analysis step.
See Example 9.5 in the Mplus User's Guide: https://statmodel.com/ugexcerpts.shtml
It depends on how you're variables are measured, especially the organizational-level moderators, your sample size, and the objective of your research. One option could be SEM analysis. If your organizational-level moderators are measured on a metric scale you can include them on your model, if they are measured as categorical or binary variable you can use them for a multigroup analysis. Depending on the sample size, the measurement model of the variables (reflective or formative), and the objective of you research you can choose between CB-SEM and PLS-SEM analysis.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism