Hi all, I am conducting a confirmatory factor analysis (CFA) on a problem behavior assessment instrument that has 7 factors. There are a total of 85 unique items and 12 of these load on more than one factor (none load on more than 2 factors). I used PROC MI in SAS to generate five imputed data sets. I imputed the missing data assuming the data were normally distributed. Because my data are ordered categorical with 3 response options, I used the Mplus to conduct the CFA using the weighted least square estimator (that uses a diagonal weight matrix with robust standard errors and mean- and variance-adjusted chi-square test statistic). I now want to use PROC MIANALYZE to combine the parameter estimates and obtain an estimate of the standard errors. Now for my question: I think I understand how the data should be set up if there were no items that double-loaded. But how do I deal with the 12 items that load on two factors each? Thanks in advance for your help. Mike. E. Michael Bohlig 1125 W. Meseto Ave. Mesa, AZ 85210 480-775-1557
