Theoretical Approaches Review on Covariance Based Sem Using Lisrel, Partial Least Based Sem Using Smart PLS and Component Based Sem Using Gesca

The aim of the research is to review theories underlying the Structural Equation Modeling (SEM) procedure based on covariance (CBSEM), partial least square (PLSSEM) and component (GESCA SEM). The methods used are meta-analysis and systematic secondary data search. Results of the study are: First, theories underlying the CBSEM, PLSSEM and GESCA SEM procedures produce different characteristics in each SEM model. CBSEM models consist of two sub models, namely 1) Factor Analysis Model consisting of a) Exploratory Factor Analysis (EFA) which is designed for a situation where the relationship between indicators and latent variables is unknown or unclear; b) Confirmatory Factor Analysis (CFA) which is used for research where the researcher already has knowledge about the structure of the underlying latent variable (construct) and c) Full Latent Variable Model (LV). 2) PLSSEM consists of two sub model, namely reflective and formative models. GESCA SEM consists of structural / inner model and measurement / outer model. Second, the primary characteristics of CBSEM, PLSSEM and GESCA SEM are requirements of the amount of data sample; the sample data origin; and the software used to calculate the data due to the different statistical formulation, namely LISREL, SmartPLS and GSCA Pro Windows. Third, the main differences among the CBSEM, PLS SEM and GESCA SEM are in the uses of the unstandardized regression coefficients (b) versus the standardized regression coefficients (β). Thus, the researchers that are going to use those procedures must consider those three important findings.