Hello, I have built a SEM model in Amos, containing interaction effects between latent construct and a dummy variable. I think I have made everything by the book:
(1) I have centered the latent constructs
(2) I have multiplied the centered constructs with the dummy variable and introduced the result in the model, as the interaction variable.
However, I get very high standard errors (huge, actually), and all the path coefficients are NOT significant (t
Analyzing moderation effects in SEM is totally different from analyzing interaction effects in traditional regression even though the purpose is the same.
Unlike in Regression, in SEM, you cannot simply multiply XM since it will affect the model convergence and increase the errors. Firstly, you need to establish the moderation effects of the said moderator variable by using the multi-group CFA. One established, you need to split the data into two group based the moderator variable you are testing. Secondly, you need to estimate the slope for the two groups separately. If the slopes between two groups differ significantly, then the effect will interact somehow along the lines. Interaction means moderation. You can download the related chapter of my book which I uploaded in RG.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling