Xu, Man posted on Monday, October 01, 2012 - 9:38 am
Dear Dr. Muthen,
I am trying to fit a second-order CFA model formed by 3 first-order factors (each of them have 3 indicators). I would like to see whether the residual correlations of these first-order factors are still statistically signficant after forming the second-order factor.
I couldn't get these residual correlations until tried to set all three of the factor loadings of the second-order factors to be 1(chi-2 n.s. when testing these loadings' equivalency). Is this a valid way to estimate the correlated residuals for the first-order factors?
It is actually a contruct that is measured 3 times. The indicators are repeated items. So some degree of measurement invairance (both loading and intercept), and correlated uniquenesses are specified for repeated indicators.
My main interest is to seperate "state" and "trait" so that I can predict both in one model using some predictor variables. I have not figured out how to specify this model in Mplus yet (there are some papers on the website but they don't include model script).
I wanted to first have a look at the residual correlations of the first-orders once the "trait" (the second-order factor) is taken into account, hence the question...
Thanks a lot!
Xu, Man posted on Monday, October 01, 2012 - 10:29 am
Sorry! I think my eyes must have been udner hullucination -I have still not managed to get the residual correlations for the first-order factors, after all.
As descrbied in the previous post, is there anyway to have look at this, or more relevantly perhaps, is it possible to specify a new variable that is based on the corresponding residual variances of the first-roder factors, so that this "state or occassion" component can be predicted or to predict other variables in a SEM model?