2
$\begingroup$

I created my first path analysis model with lavaan. Despite having a big sample (N = 600) my robust analysis has a perfect fit. I'm a bit suspicicous when something is perfect when I am doing statistics: Are the fit values plausible?. Here is the lavaan output with the fit indices:

Lavaan fit indices

Maybe relevant: My dependent variable consists of ratio values with a lot of values being 1.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

No. It's not perfect, perfect is chi-square = 0, SRMR = 0.

Look at the formulas for the various fit indices.

If your chi-square is less than your df, RMSEA will be zero.

Your CFI is actually 0.997 which is getting rounded up to 1.00.

However, your null model has 30 df, and your fitted model has 4 df - that tells me that you've estimated a lot of parameters, and that your model is pretty unparsimonious. Your fit is good, but the model that is being tested might not be as interesting as it could be.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Thx for your answer. I created my model in a way that I also can explore different relations. Thats why I estimated so many parameters. $\endgroup$
    – Rizzi
    Commented Jun 22, 2018 at 8:45
  • $\begingroup$ What would be the approach to create a more interesting model? $\endgroup$
    – Rizzi
    Commented Jun 22, 2018 at 8:57
  • $\begingroup$ A more interesting model (often) has more degrees of freedom. Sometimes an unparsimonious model is the best model you can have, because of the theory you want to test. $\endgroup$
    – Jeremy Miles
    Commented Jun 22, 2018 at 13:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.