Hi everyone, I have conducted a mediation analysis with SEM. My model fit was good but there was no significant mediation. Which aspects of this should I discuss?
Hello Dila,
testing the fit of the model will be sucessfull if the model doesn't violate the restrictions (fixed effects) in your structure. Testing path coefficients for significance is something which differs from this. If you specify a model of totally unrelated variables, the model will fit (as the model does not violate anything) but all coefficients will be non-significant. If your indirect is non-significant, there is no evidence for mediation. That's your result.
Best,
Holger
Esam Osman Yes, total and direct effect is significant.
Holger Steinmetz Thanks a lot for your answer. Do you think well fitted model is something to mention in the discussion par?
Then you report significant direct effect and insignificant mediation effect sometime the model testing using available data will show different results from the theory but in any who you report the findings from your data sets in the paper whether if confirm the theory or no
Hi all,
1) The model fit is surely something what has to be mentioned but nothing really exciting to discuss
2) Whether reporting a non-significant (direct or indirect) effect is something exciting always depends how convinced the community is/was that there is an effect. If conviction is strong, then of course reporting a "nah" is interesting and valuable. The most disadvantageous combination is "weak hypothesis" + "no sign. effect" (or even more: +bad study that is low N / erroneous measures).
HTH
Holger
- Badges
- Science topic
- Similar topics
- Philosophy
- Moral Philosophy
- Free Will