Robust ML for CFA or SEM
Hi there. Is it possible in JASP to select a robust ML estimator (MLM or MLR) for CFA or SEM?
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Categories
Supported by
Comments
CFA:
SEM:
Hi. I tried these options, but except for SRMR, the fit results largely do not change and do not match the robust values in R. Below are the RMSEA values in R using "MLM" as the estimator. Normal ML RMSEA is 0.063 and the robust values are 0.05 or 0.054, depending on which robust values are used.
The CFA RMSEA values in JASP with the standard error set on "standard" is: (this matches the normal ML)
But when changing it to "robust" the RMSEA results do not change.
However, the R console in JASP gives the correct information:
What if you change model test in the SEM analysis (see the second screenshot)? Also what if you pick mimic/emulate Mplus?
Hi both, as per the following feature request, I would hazard a guess that the options you require are not currently available, but it does look as if a team member has been issuing updates to the required module so that they will soon be available: [Feature Request]: Robust Estimation in CFA · Issue #2162 · jasp-stats/jasp-issues (github.com)
I had apparently inadvertently requested part of the same thing because a recent study of best practices for CFA has noted that robust fit indices would be very useful thing to have added.
[Feature Request]: add robust and t-size fits indices for CFA and SEM · Issue #2885 · jasp-stats/jasp-issues (github.com)
In short, not yet, but it should be possible soon.
I have no idea however why this does work correctly in the console. Maybe because you can access every feature of the underlying library if you run the code yourself but those options are not yet available in the graphical menus?
Note:
The only thing that changes if you click robust in the CFA menu is the standard error. Actually you want to choose an entirely different estimator, which is not available yet by the menu. This will also give different fit indices et cetera
TarandeepKang, this is fantastic news. Looking forward to the update, as this will help a lot with my students.