Solved – Doing CFA on a known theoretical model, but having problems with convergent and discriminant validity

amosconfirmatory-factorpsychometricsstructural-equation-modelingvalidity

I did a survey based on the UTAUT model / theory with standard questions used. The N was 150. I am trying to validate the model for my survey with structural equation modeling (via the AMOS software). But I have heard that before testing the model fit, "it is absolutely necessary to establish convergent and discriminant validity, as well as reliability, when doing a CFA" (confirmatory factor analysis).

Doing CFA, I found out that factor loadings are OK (using AMOS, loadings > 0,7).
I did CR for reliability testing, everything OK – over 0,7. But (not like other published papers) I have problems with convergent and discriminant validity (note results below).

enter image description here

Is there some way to deal with this or do I have to restate the questions and repeat the survey? (I think doing exploratory factor analysis is not logical if so many others have proven the theory works, so does that mean something must be wrong with my data?)

Best Answer

Basically you are doomed. But you may llok at Zaid and Bartea, 2011 to calculate discriminant validity after Segars, 1997. It involves calculating fit coefficients with two constructs at a time twice having the first one correlated and the second uncorrelated between constructs. Comparing the fit measures between these two methods if they are significantly different as per CMIN criteria for 1 degree of freedom, then discriminant validity is said to exist.

Related Solutions

Solved – Can exploratory, and not confirmatory, factor analysis be used on dependent variables

EFA can be conducted on dependent variables. Then you can use the factor saved scores in the regression.
Often practical data do not commensurate with the theoretical factors. So, you could also try an EFA separately, that is, try both. See if the results match.

Please see these links:

Difference between exploratory and confirmatory factor analysis in determining construct independence

Differences on exploratory factor analysis, confirmatory factor analysis and principal component analysis

SPSS Amos – How to Create Item Parcels in SPSS Amos

I've got a few notes here on the topic of item parcelling.

But Jeremy Miles makes the main point. You don't parcel in Amos. You use other software (for Amos users, this is typically SPSS) to combine items into parcels. There are a variety of ways of making parcels, but common approaches simply involve dividing the items in a set into parcels and taking the sum of mean of items within parcels to form the parcels.

In SPSS, you can often do this efficiently by using syntax. E.g., Something like this:

compute parcel1 = mean(i1, i2, i3).
compute parcel2 = mean(i4, i5, i6).
execute.

Best Answer

Related Solutions

Solved – Can exploratory, and not confirmatory, factor analysis be used on dependent variables

SPSS Amos – How to Create Item Parcels in SPSS Amos

Related Question