Get a Relatively Good Model
Construct a Model in Amos
Don't try to memorize the process. It's pointless:
CMIN --- CMIN/DF: < 3 (good); < 5 (OK) CMIN --- P: > .05 RMR,GFI --- GFI: >.95 RMR,GFI --- AGFI: >.80 Baseline Comparison --- CFI: >.95 (good); > .90 (OK); > .80 (barely OK) Parsimony-Adjusted Measure --- PCFI: >.80 RMSEA --- RMSEA: <.05 (good); >.05 && <.10 (OK); >.10 (barely OK) RMSEA --- PCLOSE: >.05
CMIN/DF and P-value together tell you whether you model has a good fit; if the result is not within the range of metrics while you have a big number of sample, it is OK.
Remove some variables with low loadings.
Go to Modification Indices >>> Covariances and look for big number of M.I. between each pair of errors. Covariate ONLY those pair of errors on the same factor.
Go to Estimates >>> Matrices >>> Standard Residue Covariances and look for big numbers above or close to 2.58. Remove the variable that causes problems.
Validity and Reliability
Metrics (for every factor):
CR > .70 AVE > .50 MSV < AVE ASV < AVE
If you have convergent validity issues, then your variables do not correlate well with each other within their parent factor. Stop using CFA.
- Common Method Bias
- Measurement Model Invariance
For more information, please see Dr. Gaskin's website.