Dear everyone,
When I analyze and compare an identical model between groups, I always conduct measurement invariance analysis (Byrne, 2004) to demonstrate all participants interpreted the survey questions in a similar manner, which allows me to compare groups (e.g., means, path coefficients).
When I compare my model between groups, I found a significant indirect effect of variable A on variable C through variable B in group 1 (Standardized B = .161; SE = .036; 95%CI [.098; .238]) and in group 2 (Standardized B = .082; SE = .021; 95%CI [.047; .129]).
In this context, how do I know if the indirect effect observed in group 1 is significantly greater than the indirect effect observed in group 2? In the same veine, how do I know if the path coefficient observed from variable A to variable B is significantly different between groups, although it is significant (or not) within both groups.
Thank you for your help.
Robin
I guess through different classroom activity methods that show and permit that measurement of each group. We can always change the paradigms.
Measurement invariance tells you whether the data fits one group more than the other. More or less, it is equivalent to differential item functioning. If you want to compare groups in SEM, you should conduct a multi-group analysis as Ali H. Al-Hoorie has already stated.
I guess a straightforward answer to your question can be found in the CIs of the two effects. The values you mentioned indicated large overlap between the two CIs.
You can do it with multigroup analysis in AMOS using the CMIN invariance test for nested models. Here is an example:
https://www.youtube.com/watch?v=w5ikoIgTIc0
Regarding the indirect effect. If the two direct effects are statistically different according to the test above, it is sound to accept that the indirect effect is also different. I do not think any reviewer would disagree with this reasoning. However, it is important to first test if the indirect effect is statistically significant. This you would do it using the bootstrap test for indirect effects. You can do similar things with R and other SEM softwares, but I do not know the procedures in these other programs. Sorry. I hope have been of help.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation