Xu, Man posted on Monday, December 02, 2013 - 1:12 pm
Dear Dr. Muthen,
I want to compare the regression coefficient between the sum score method and latent variable approach, using the simulations.
The population model is latent variables with an observed co variate. The correlation of the latent and the covariate is the interest of study. I wanted to make the analytic model to be based on sum scores of the simulated items. I don't think the command DEFINE can be used to create observed sum score. So I tried to constrain all factor loadings to 1 and item residual to 0 for all items. However, the analysis doesn't run (no convergence). So I don't think this works for creating the sum score.
Is there a way to analyse the simulated items as sum scores within the MPlus?