As per the attachment, my research focuses on the mediators effect on the relationship between the IV and DV. Based on EFA, I found that IV1 and IV2 are the subsets of IV whereas DV1 and DV2 are subsets of DV. I would like to predict DV based on IV, mediator and demographics of sample. In the case of IV, do I select total IV based on EFA findings or select IV1 or IV2 ? Or do I need CFA to validate the choice of one factor or two model based on goodness of fit? What else could you suggest?
according the statement, IV1 seems IV2 are muture exclusive, i.e individule subject either have iv1 or iv2, not both. If yes, you have the following rregression,
same interception and different slopes (for Iv1 and Iv1
dv=intercept + ivstat*IV, ivstat=1, if iv=iv1, ivstat=0, if iv=iv2
Or different interception different slopes;
dv=intercept+ ivstat+ ivstat*IV
ivsta*IV menas the interaction between ivstat and IV
One typo: the equatoins should be:
dv=intercept +IV+ ivstat*IV, ivstat=1, if iv=iv1, ivstat=0, if iv=iv2
Or different interception different slopes;
dv=intercept+ ivstat+IV+ ivstat*IV
I think you need a structural equations model here, with the subsets of your IV and DV articulated. MPlus will model your diagram pretty effectively.
Hi Chan,
I think depend on what & how you structured your actual research objective, question & hypotheses.
1) if your research objective is merely investigating the high level mediator effects of IV-DV, perhaps you can treat respective IV & DV as a whole construct i.e. inclusive its sub-components and use Baron & Kenny mediation analysis
2) If your literature review, research objectives & questions, hypotheses dive into the sub-components IV1, IV2, DV1 & DV2 as well, perhaps you can use SEM with first- or / and second-order latent variables - when you use SEM, it comes with CFA
All the best.
Regards,
Fung
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression