Hi I am analyzing data from a RCT (n=151)with cluster randomization (only 4 clusters), and have chosen to control for thos effect by running av COMPLEX type. I have used only change scores calculated in SPSS first. The fit is OK, but drops when I change 1 single outcome variable (4 in total), I don't understand why:
RMSEA from.041 to .079 CFI from .97 til .89 TLI from .94 to .78 SRMR from .053 til .056 chi square from 5.1 to 5.26 (p=.000) SRMR from .053 to .056
All correlations are roughly the same between variables in boht models except for one that is more logical in the "bad model"