Hello everyone!
I have a hypothesized model with two mediators. The fit indexes are good in this model and the indirect effects are significant as expected. I would also like to test an alternative (reverse) model to test the validity of the original model. However, the fit indices of the alternative model shows the model has poor fit. I have concerns as to whether I can interpret mediation effects in this case. It would not be correct to interpret the mediation effects in non-fit models, would it?
Thank you very much in advance.
Hello Ceren!
In my opinion, it would not be advisable to interpret the mediation coefficients in a model without a good fit to the data, as they are usually based on the estimates provided by the other effects. As such, the same rules should apply. If your reversed model does not fit the data well, it is an indication that you cannot reproduce your input matrix adequately, and as such, your estimates are not well based on your data. Consequently, it is not advisable, in my opinion, to interpret those estimates as such. However, It is possible to interpret the poor fit of the reversed model as evidence supporting your "forward" model. This may do the trick if your interest is in providing evidence for the direction of your original model, making it rather unnecessary to discuss the reversed mediation effects anyway.
I hope it helps,
Best wishes!
Dear Cristóbal Hernández,
Thank you very much for your comments and explanations. My idea is so similar to yours. It is necessary to consider fit indices as a prerequisite and not to interpret indirect effects. Of course, this situation might confirm our hypothesized model rather than the alternative model.
Best!
- Badges
- Science topic
- Similar topics
- Psychology