RMSEA too high? Problems with this fit index.
HTML-код
- Опубликовано: 11 сен 2024
- Together with CFI and SRMR, the RMSEA is one of the most popular fit indices for path analysis, CFA (confirmatory factor analysis) and SEM (structural equation modelling).
Unfortunately, the RMSEA can give seriously misleading answers to the question whether a model has an adequate fit in models with small df. Primarily, this is a problem for path analyses and simple CFA models where a small number of degrees of freedom is quite common. In that case you can get a RMSEA over 0.1 even for a correct model. A combination of high RMSEA and low SRMR is then also possible.
This video explains the problem with the RMSEA in models with small df and shows possible solutions.
How to book a statistics CONSULTATION for LAVAAN (SEM, CFA or path analysis):
www.regorz-stat...
To my other tutorials:
www.regorz-stat...
Based on:
Kenny, D. A., Kaniskan, B., & McCoach, D. B. (2015). The performance of RMSEA in models with small degrees of freedom. Sociological methods & research, 44(3), 486-507. doi.org/10.117...
You have just saved me a ton of time, and your calm voice has lowered my blood pressure -- thank you!
Thanks for explaining a complex issue in such a simplified manner. You helped me in a big way.
Great video! Thank you so much.
THANK YOU! Your video is very useful.
Hello, what number ranges does the small RMSEA value refer to? Also, how many people does a low N value include? 100? 200? Thank you for the video. You have been very helpful.
@@hamdiozturk-c2h For that you should look into Kenny, Kaniskan, & McCoach (2015), there are tables in their article.
Thank you!!!
how do you know if your degrees of freedom is too small?
That depends on the sample size. In the journal article linked in the video description (Kenny et al., 2015) you can find results from simulation analysis for different combinations of degrees of freedom and sample size.
rmsea in my model is coming out to be 0.082 is their a way I can reduce it to acceptable limit
First, it depends on the degrees of freedom (df). With very low df, the RMSEA isn't really relevant - that is the point of the video.
If that is not the case, you could use modification indices to assess which additional freed parameters could improve the model fit. Of course, you should only free those additional parameters that make sense from a theoretical viewpoint.
I have RMSEA of 0.142, with degrees of freedom of 4, and sample size of 640. How can I fix it because Kenny et al. (2015) said the issue is only for small df AND sample size. I don't think I can call my sample size small and yet I still get bad RMSEA eventhough CFI and SRMR are both good
In that case I would look at modification indices in order to check whether there are necessary model changes.
What are the bad fit indices of goodness of fit?
I am afraid I don't understand your question.