JASP Omega different from RStudio

Frans

I get an Omega value of 0.708 in JASP with the function Unidimensionality Reliability, but a value of 0.74 in RStudio.

I can't see the R code for this function in JASP, but the R Code I used in omega(df.items, nfactors=1). I attached the 4-item dataset.

Dataset.csv

JASP Omega results.jasp

julius

Hi,

the difference stems from the psych package in R using the data correlation matrix to find the factor loadings to compute omega, and in JASP we use the covariance matrix, because we deem it more informative.

Julius

Frans

That makes sense. Thank you!

patc3

So basically one is standardized and the other isn't?

Frans

Yes, that's how I understand it.

One can add the following R code to give the unstandardized (from covariance matrix): covar = TRUE.

To be exact, to get the same value as JASP: omega(Dataset,  nfactors = 1,  fm = "ml", covar = TRUE).

julius

Exactly. We have had a discussion with Klaas Sijtsma about this, and we came to the conclusion that through the standardisation differences in the variances are lost. It is true however, that we offer standardised alpha - because it is somewhat of a convention - and in the Bayesian version we offer all coefficients standardized. So eventually, we should probably also offer all coefficients standardized for the frequentist reliability as well.