SEM in JASP, need help with AVE (beginner)

caro96

Hello everybody!

This is my first time working with JASP as part of a seminar paper and I don't have much prior statistical knowledge. So far I have done a confirmatory factor analysis and a regression using SEM (lavaan). As a measure of validity we are supposed to report the Composite Reliability and the Average Variance Extracted, among others. Unfortunately, I have not found a function for this in JASP yet. When I looked at the Factor Covariances table, I noticed that only the covariances of the independent variables were measured and the dependent variable is not listed. How can I change this? Ideally, I would like to have a correlation table that also maps the Composite Reliability and the Average Variance Extracted.

Thanks a lot in advance!

Best regards

Carolin

JMBKoch

Hi Carolin,

If you use a SEM model to regress latent variables (factor that you modeled yourself) on one another, you don't actually need to assess their reliability in my opinion. Composite reliability estimates such as cronbach's alpha are used to gauche whether it's wise to combine a set of items into a sum- or mean score. Since you use latent variables rather than a sum- or mean score, there is no need for this step. In fact, the SEM module already takes the measurement error into account by modeling it. In a way, computing a reliability estimate is then redundant. Of course, in practice reviewers, collaborator or supervisors may disagree. In that case you can get estimates of composite reliability (e.g. Cronbachs alpha) for a set of item by using JASP's Reliability module. To my knowledge, computing average variance extracted is not possible in JASP as of today, but I would also not suggest reporting this value with a confirmatory factor analysis either.

I'm not 100% sure I caught your question right about the covariances? What do you mean by independent and dependent variables? Are you talking about the variables in your measurement model, or the ones in your structural (regression) model? In the factor covariances, there should be the covariances tables between all latent variables.

Best,

Michael

caro96

Hi Michael,

thanks a lot for your help! Since our supervisors of the seminar paper want to have these values, I have now simply calculated the AVE and CR using Excel. 

The covariance table now also has the correct format. The only thing that still bothers me are the estimates of the coefficients. For example, the values "-2.007e-5" and "-6.742e-4" were calculated and unfortunately I don't know how to convert them into a number. Maybe you can help me with that?

Thanks a lot in advance!

Kind regards

Carolin