In this work, we show how to use Regularized Generalized Canonical Correlation Analysis (RGCCA) in structural equation modeling with latent and/or emergent variables. This new approach produces consistent and asymptotically normal estimators of the parameters. RGCCA relies on a well-grounded optimization problem and the global convergence of the algorithm used to solve this problem is guaranteed. We also propose a maximum likelihood (ML) estimation method for estimating the parameters of the model. RGCCA and ML are evaluated in a Monte Carlo simulation and lead to similar results. RGCCA and ML are also compared on the ECSI data for the mobile phone industry and produce very close results.