Hi,

the AVE (like other fit *measures*) were invented because people did not like what the chisquare *test* said about their models ;) And if it your test is significiant then either the data assumptions are so heavily violated that the test does not work properly (hence, if e.g. your data is heavily nonnormal, use a corrected version of the chisquare test) or that the model structure is misspecified (which in my estmation is the case in 90% of models). This misspecification may be trivial, which means that your overall structure is ok but you missed some things and these omissions would not bias the effects of the correct part. However, it may be fundamental and the overall structure is wrong. The problem is, you don't know what's the case.

Unfortunately, AMOS does not help you in this regard, hence, your review may be the necessary input and incentive to now invest a few days and use the free lavaan package in R (www.lavaan.org).

If you allow me a self-citation, this paper (study I) presents a CFA that was initially misspecified and it presents what I did to locate and adress the problem. The reason why I post this is that this is the only published case I know where reviewers did not cry in agony when an author trims the model (a weird collective habit in our field) and I was "allowed" (so to speak) to demonstrate the process. In the paper, we also discuss that this model trimming has (of course) some negative implications. However, in my view, an acceptable cost in contrast to keeping a perhaps fundamentally wrong model.

Rosman, T., Kerwer, M., Steinmetz, H., Chasiotis, A., Wedderhoff, O., Betsch, C., & Bosnjak, M. (2021). Will COVID‐19‐related economic worries superimpose health worries, reducing nonpharmaceutical intervention acceptance in Germany? A prospective pre‐registered study. International Journal of Psychology. https://doi.org/doi.org/10.1002/ijop.12753

There is an available R code associated with this paper that you can download here:

https://www.psycharchives.org/handle/20.500.12034/4092

Hope that helps

Holger

In order to establish discriminant validity there is need for an appropriate AVE (Average Variance Extracted) analysis. In an AVE analysis, we test to see if the square root of every AVE value belonging to each latent construct is much larger than any correlation among any pair of latent constructs.