Seminar in Psychometrics

COMPS seminar main page

Old and new approaches for including interaction and quadratic terms in structural equation modeling

Date and time: February 27, 2023 (4:00 PM CET)
Place ICS CAS room 318, Pod Vodárenskou věží 2, Prague 8.

Abstract. In a linear regression model with multiple predictors, it is trivial to extend the model by adding interaction terms or quadratic terms. But when the predictors are latent variables, this is far from trivial. In this presentation, I will give an overview of several approaches that have been proposed in the literature to extend linear structural equation models (SEMs) to so-called nonlinear structural equation models (NSEMs) that include quadratic or interaction terms involving latent variables.
I will discuss the latent moderated structural equations approach (LMS; Klein & Moosbrugger, 2000), the nonlinear structural equation mixture approach (NSEMM; Kelava & Brandt, 2014), and several variants of the product indicator (PI) approach (Kenny & Judd, 1984; Marsh, Wen, & Hau, 2004). All these methods use a system-wide estimation approach and estimate the free parameters of the model simultaneously. This is in contrast to the 2-stage method of moments estimator (2SMM; Wall & Amemiya, 2003) where factor scores are computed in a first stage, and an errors-in-variable regression approach is used in the second stage. In this presentation, I will also propose an alternative approach that is similar in spirit to 2SMM, but where we avoid the explicit calculation of factor scores. The approach builds on the (local) structural-after-measurement (SAM) approach that was recently proposed by Rosseel & Loh (2022).

References.
Klein, A.G. & Moosbrugger, H. (2000). Maximum likelihood estimation of latent interaction effects with the LMS method. Psychometrika, 65, 457-474.
Kelava, A. & Brandt, H. (2014). A general nonlinear multilevel structural equation mixture model. Frontiers in Psychology, 5, 748.
Kenny, D. Judd, C.M. (1984). Estimating the nonlinear and interactive effects of latent variables. Psychological Bulletin 96, 201-210.
Marsh, H.W., Wen, Z., & Hau, K.-T. (2004). Structural equation models of latent interactions: evaluation of alternative estimation strategies and indicator construction. Psychological Methods 9, 275-300.
Wall, M.M. & Amemiya, Y. (2000). Estimation for polynomial structural equation models. Journal of the American Statistical Association, 95, 929-940.
Rosseel, Y. & Loh, W.W. (2022). A structural after measurement (SAM) approach to structural equation modeling. Psychological Methods. Advance online publication. https://dx.doi.org/10.1037/met0000503