Comparison of methods for estimating and testing latent variable interactions

Structural equation modeling methods for estimating and testing hypotheses about an interaction between continuous variables were investigated. The methods were (a) Bollen's (1996) 2-stage least squares (TSLS) method, Ping's (1996) 2-step maximum likelihood (ML) method, and Jaccard and Wan's (1995) NE method for the Kenny-Judd model (Kenny Judd, 1984); (b) a 2-step NE procedure and ML estimation of the Joreskog-Yang model (Joreskog Yang 1996); and (c) ML estimation of a revised Joreskog-Yang model. The TSLS procedure exhibited more bias and lower power than the other methods. Under ML estimation of the Joreskog-Yang model, Type I error rates were not well controlled when robust standard errors were used. Among the remaining procedures, the Jaccard-Wan procedure and ML estimation of the revised Joreskog-Yang procedure were most effective, with the latter having some small advantages over the former.