I'm not sure that SPSS has an option to treat the factor indicators as categorical. And I don't know much about EFA in SPSS. For categorical outcomes and EFA, Mplus uses weighted least squares estimation. Both varimax and promax rotations are used.
Hi, I have some questions concerning "odd" results in EFA in Mplus. I recently computed a 3-factor model in Mplus and SPSS (for cross-checking) with 7 continuous indicators (same data, identical model specification and number of cases, both oblimin rotation). In SPSS, everythings fits the 3-factor structure well, with only negligible cross-loadings. In Mplus, one item has a large cross-loading that even exceeds the loading on the factor it is supposed to belong to. Also, the other indicator of the same factor has a loading >1 and a negative residual variance (although there is no error message in the output). Reducing the number of factors does not seem to be a solution because the 1- and 2-factor solutions both have a very poor fit. What am I supposed to do about this? Also, can you tell me how Mplus obtains 3 degrees of freedom for the EFA? What are the underlying parameter restrictions?