Reply With QuoteWhat is the RMSEA value like? df?
Hey,
I performed a multilevel cfa with 320 cases at the first and 37 cases at the seocond level. The fit indices ar good (NNFi, CFI etc) except for the SRMR, which is much too high (about 0,24). I wonder what the underlying reason is and whether there is a possibility to reduce thta SRMR index. Thanks a lot for any suggestions.
Jana
What is the RMSEA value like? df?
I am not sure what software you are using. In Mplus it has Modification Indices which suggest possible problems with your model (although you then have to figure out exactly how to respecify it to deal with the problems which will vary with type of model).
Are some of your residuals (disturbance terms) or factor loadings on certain paths especially low (or violate your theory)?
If everyone one of your model fit test but one are good, you might not actually have a problem (this is most true with chi square, but I think includes others as well).
i'd try and bootstrap the hell out of it & see what happens... i happened to go to yet another thesis defence specifically addressed to see how fit indeces behave in prescence of several types of model misspecification in CFA and, as it turns out, the SRMR is sensitive to model misspecification, especially missing factor covariances... which tend to get lost along the levels of multilevel models, but, again, this is just a hypothesis. knowing more about your data would help...
Dason on the Cauchy distribution:
"YOU BETTER LOOK OUT BECAUSE THIS IS SOMETHING THAT IS GOING TO GET YOU"
bumping..........
Why is this thread being bumped (because the answer was useful to others or because more comments are desired)?
"Facts are stubborn things, but statistics are more pliable." Mark Twain
Posting Permissions
|
|