ABSTRACT
This chapter provides a largely nontechnical description of an alternative technique to include interactions of latent variables in structural equation models. It also provides a generic description of a model with interactions of latent variables with multiple indicators. The chapter presents the two-stage least squares (2SLS) method to model such interactions. It discusses the D. A. Kenny and C. M. Judd and related methods for handling such interactions. If the new measurement equations and nonlinear constraints are included, Kenny and Judd showed that the coefficients for the nonlinear variables could be consistently estimated with generalized least squares estimation. The chapter compares the two techniques with the simulation and empirical data that Kenny and Judd provided in their original paper. It focuses on a 2SLS method to handle interactions of latent variables that have multiple indicators. The chapter concludes with comments and contrasts the alternative methods and their properties.
