I am a PhD scholar. When importing the rotated component matrix in Amos 23, I'm having trouble. My factor loads while conducting CFA have different values than the loadings I get in SPSS after performing factor analysis and transferring them to Amos 23 using the pattern matrix plugin.
I have an acceptable model fit, but because of fluctuation in cfa factor loadings, my AVE falls below the.5 cutoff. Is there a way to fix this?
I would say it is generally expected that you would get different factor loadings when conducting CFA in AMOS versus exploratory factor analysis (EFA) in SPSS. CFA and EFA are different approaches/models that can lead to different results for a variety of reasons, for example, because they may imply different assumptions/restrictions regarding the covariance/correlation structure of the variables.
Your loadings are different because EFA and CFA are entirely different procedures which answer different questions and (generally) use different estimation methods.
One approach you could try here is to run your EFA using maximum likelihood as the method of extraction - doing this may make your loadings a little more similar. The downside is that it tends to create Heywood cases more often than not. If you do go down this path, I'm not 100% certain that SPSS will alert you to this issue so make sure to carefully check how much of the variance your factors are accounting for in your observed variables - if you see one exceed 100%... you have a problem.
Hi,
if your factor loadings are low in a CFA model, one plausible option is that the model structure is simply misspecified and the loadings are biased.
What is the chisquare test, df, and N of the model?
Best,
Holger
Holger Steinmetz David Hicks Christian Geiser although the CR is higher than its threshold will value of AVE make any difference then.
As suggested in the below recommendation
Nashwan AL-Emad , literature suggest 0.6 as the minimum acceptable value for Cronbach Alpha
Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2009). Multivariate Data Analysis (7th ed.). Upper Saddle River, New Jersey: Pearson Education Limited.
Average Variance Extracted (AVE) is higher than 0.5 but we can accept 0.4.
Because Fornell and Larcker said that if AVE is less than 0.5,
but composite reliability is higher than 0.6, the convergent
validity of the construct is still adequate (Fornell & Larcker, 1981)
Hi,
just to be sure: The chisquare ist 1 thousand, 6 hundreds etc? (asking because of your use of the decimal points). If yes, you model is highly misspecified. The huge degrees of freedom suggests that you have a very large number of indicators. Can you tell how many indicators and how many factors?
I would not interpret the loadings and go back to the drawing board. Select 2-3 indicators which are crystal clear (from the perspective that these should reflect the same underlying latent).
Here is a paper, in which we also had a misspecified model and it gives some clues how we dealt with that:
Rosman, T., Kerwer, M., Steinmetz, H., Chasiotis, A., Wedderhoff, O., Betsch, C., & Bosnjak, M. (2021). Will COVID-19-related economic worries superimpose virus-related worries, reducing nonpharmaceutical intervention acceptance in Germany? A prospective pre-registered study. International Journal of Psychology. https://doi.org/10.1002/ijop.12753
Article Will COVID-19-related economic worries superimpose health wo...
Rcode
https://psycharchives.org/en/item/abbc99f1-617e-41eb-a167-32f3792c57df
Good luck
Holger
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication