Omega > 1

[Rizqy Amelia Zein]

Hello all,

I'm doing a CFA for social dominance orientation (only dominance pro-trait and contra-trait, leaving off egalitarianism) scale and found bizarre stuff when examining its hierarchical omega using omegaFromSem() function. It says the omega total equals 1.33 while its hierarchical omega is 2.29. Since those values are implausible, could anyone explain why this happened?

For a context, I used a MLR estimator and the model seems to converge well (X2(21)=1281.019, p=.000, robust RMSEA=0.048, 90% CI RMSEA [0.034, 0.028], CFI=0.975, TLI=0.959, SRMR=0.035).

I also examined its composite reliability, and despite being not super great, the values look normal (pro-trait = 0.764, contra-trait = 0.648).

I genuinely appreciate any leads, and thanks a lot in advance!

Best

Amelia

[Terrence Jorgensen]

found bizarre stuff when examining its hierarchical omega using omegaFromSem() function. It says the omega total equals 1.33 while its hierarchical omega is 2.29. Since those values are implausible, could anyone explain why this happened?

That function is from the psych package, so you would have to ask its maintainer (providing data and script for a reprex).

If you install the semTools package, you could instead use the compRelSEM() function to obtain omega.  

• By default, it returns "hierarchical" for each factor in your model (so only pay attention to the factor of interest, or use the omit.factors= argument to suppress ones not of interest).  
• You can obtain omega "total" by setting return.total=TRUE, which assumes that you are creating a composite from ALL indicators in your model (unless you omit any using the omit.indicators= argument).

The way I designed the compRelSEM() function, I doubt values > 1 are possible.  Perhaps your measurement model has cross-loadings?  I think that might make values > 1 possible when using the original formula based on simple structure, which might be what the psych package assumes.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

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

http://www.uva.nl/profile/t.d.jorgensen

[Rizqy Amelia Zein]

Hi Terrence,

Thanks so much for taking the time to answer my question. I included return.total=TRUE and the result is different from what omegaFromSEM() gave. I will get in touch with the psych maintainer.

Again, thank you!

Amelia

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/03bded10-ba32-4bb5-acf9-e74c6a9999a9n%40googlegroups.com.

--

Rizqy Amelia Zein

http://psikologi.unair.ac.id/

[Terrence Jorgensen]

If you provide your syntax with explanation about which are general or method factors, as well as what composite of items (e.g., (sub)scale means) you want to know reliability of, I could at least tell you whether compRelSEM() is providing what you are expecting.