Structural Equation Modeling is a common technique applies to study the causal relationship between constructs or latent variables. It is based on Regression Analysis and Factor Analysis (see Hair et al. 2007), Basically, there are two type of SEM, i.e. covariance-based SEM (for example LISREL and AMOS belonged to this type) and variance-based SEM (for example SmartPLS).
In my expeperinces using SEM, the most difficult task is to determine whether your latent variable that is can not measured directly, a formative of reflective construct. For example, does Customer Satisfaction is formed by product's price, products' brand, etc. OR reflected on product's price, products' brand, etc. Recently, some techniques have been developed to answer this question (Tetrad Analysis is an example).
In addition the good work showed by Dr. Janarthanan and Dr. Tsiotsou, you can browse the internet for searching another literature.