Attached please find one paper written by Li et al. (1998) that compared three approaches of testing moderation by SEM (Joreskog and Yang, 1996; Jaccard and Wan, 1995; Ping, 1996). While among them, I preferred Ping’s 2-stage approach. This approach suggested that the parameters in a linear-latent-variable-only measurement model are estimated first and then lambdas together with the error variances are used to calculating the values of interaction. Finally the whole structural model will be specified by fixing the loading and error variance for each indicator. All the programs are in the appendix of the paper by Li et al. (1998).
There are at least three different methods of testing moderation by SEM. This paper compared some of them with the same example. You can find the differences in each program attached in the appendix of this paper. If u are interested, we can have a discuss~2#shenym