For categorical items using WLSMV I am running a multilevel factor model however there are some limitations to testing nested models, identifying sources of mispecification (etc) in a multilevel analysis and I wanted to output a within-level polychoric correaltion matrix and an asymptotic covariance matrix of the polychoric correlations. This is straignforward enough however where I would like some guidance is in regard to:
1. whether I should be using the polychoric correaltions matrix or the asymptotic covariance matrix as input for subsequent analysis, my thoughts are the latter; and
2. the most appropriate choice of estimator for the analysis of this within-level matrix.
I am trying to run a confirmatory tetrad analysis (CTA) with ordinal indicators using the CTANEST1.SAS macro developed by Hipp, Bauer, & Bollen (2005). According to the instructions, the way to get the asymptotic covariance matrix of the polychoric correlations using Mplus is to run a saturated model using the WLS estimator model with no latent variables and allowing all the indicators to co-vary and then requesting and/or saving the Tech 3 output. These instructions (for Mplus version 2.1) were presumably provided by Dr. Muthen. However, everything Iíve read on the Discussion Board seems to indicate that you canít save the ACM using Mplus. Could you please clarify? Thank you very much.