CFA is great with RMSEA 0.059, CFI .932 GFI .900 ChiSq 3.082 df 278.
However, in structural model, notes as below:
The following variances are negative. (Group number 1 - Default model)
e44 e45
-.350 -5.141
Hello,
Large negative residual variance (like in case of e45) can be a sign that your model is not appropriate for your data and needs to be changed.
Possible reasons: 1. Small sample size, 2. model misspecification, 3. very skewed variables (floor effects). 4. Outliers or influentual cases.
You have just-identified model so "2' is not your case.
Hello,
Just one small addition. CFI of .932 is acceptable but good model fit starts with CFI between .950 and 1. Also sometimes small chi-square stems from small sample size (however chi-square is a very crude test and as such should be always reported but the model should not be disqualified based on chi-square exclusively).
Best wishes
Hello Mohd, in addition to the valuable suggestions already given, you may check whether there are only two indicators measuring a construct. In this case there may be an identification problem resulting in a (standardized) factor loading > 1.0 and a negative error variance.
Regards,
Karin
Thank you Marina, Witold and Karin.
My sample size is 599.
However, after corrections made, my result turned to:
GFI .890, CFI .918 RMSEA .068 ChiSq 3.746
But
The following variances are negative. (Group number 1 - Default model)
e36
-2.502
What is the explanation for that? What should i do?
Hello Mohd, try to analyze parts of your model separately in order to trace the origin of the problem. From your model fit criteria it seems to me that the model does not fit too well. Smaller models may give you a hint where the misspecification is located.
Regards, Karin
If your model is just-identified, your chi-square should be 0 and the fit should be perfect. So how do you get chi-sq higher then 0? It is not clear for me...
Thank you all. Will do as Karin suggest. Will get back later.
Hello Mohd,
Did the problem solve after analyzing the models separately?