LikeLike

]]>LikeLike

]]>Just to clarify, what I was referring to is that your “multipleMediation” model from above and my own model both result in perfect fit(?): CFI = TFI =1, RMSEA = SRMR = 0, p-value that RMSEA <= 0.05 is NA. It is true, your model estimates AIC & BIC (in my model, these are missing for whatever reason). Perhaps, I was wrong to assume that this is due to the model being just identified. I was mainly assuming this, because if the covariance of the two mediators is not estimated, the fit measures become more meaningful both in your example and in my own model.

Disclaimer, I haven't yet looked into the papers you recommend.

LikeLike

]]>LikeLike

]]>LikeLike

]]>Your multiple mediator model example seems to be just identified (has 0 degrees of freedom) and thus does not estimate model fit. The same happens with my own two mediator model, which is essentially identical but has a few control variables added. I could live with this, if I knew it was a necessary property of multiple mediator models (with no latent variables). Unfortunately, I could not confirm this anywhere. Indeed, Appendix B in the above cited Preacher & Hayes 2008 paper has an SPSS output which provides fit statistics.

So, I wonder if you could give me any pointers whether I should be worried about the lack of model fit statistics and more broadly, why some multiple mediator SEM models are just identified and others are not.

It’s much appreciated.

Alexander

LikeLike

]]>LikeLike

]]>LikeLike

]]>