Interpretation CFA output

[Saskia De Veth]

Hi!

I developed a scale containing four factors. 

In R the CFA provides the output as shown in the attachment.

Can you help me with the interpretation of the values? Which are important and what do they say about my model?

The Z-values seem okay? All above or below 2/-2. The RMSEA <0.08 but the CFI and TLI are not that high (high enough?). What does this mean?

Thank you so much!

[Jeremy Miles]

Your items are not very highly correlated - this might mean that they are measured with error, or it might mean that they are not measuring similar constructs.

That means that, by things like chi-square and RMSEA, your model is not very bad - but the trouble is that no models are especially bad. Try running a one factor model, it won't be very bad. In fact the null model, which proposes that there are no factors at all, is not very bad - your fitted model only reduces the chi-square by about half.

Your chi-square and RMSEA are good, because you don't have the power to find bad fit. That's why you should always look at CFI/TLI as well. (The reverse is true - very high CFI combined with poorer RMSEA/chi-square tells you that you had a lot of power. A conclusion you can draw is that CFI always works.)

Jeremy 

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[Saskia De Veth]

Hello Jeremy,

Thank you for your response. Is there also a conclusion I can draw from the z-values and p-values of the latent variables and the covariates? Do the negative numbers need to be reverse scored? And do the numbers say something about their additive value to the model?

Hope you can help me out..

Thankyou!

Op zaterdag 21 mei 2016 19:27:17 UTC+2 schreef Jeremy Miles:

[Terrence Jorgensen]

Hi Saskia,

You have very general questions that are not software-specific, so SEMNET would be a more appropriate forum.  It has a wider audience of SEM experts, so response time is usually shorter.

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

UvA web page: http://www.uva.nl/profile/t.d.jorgensen