Structural equation modeling (SEM) is a statistical modeling technique that is very cross-sectional, linear and general. Included in this SEM is factor analysis (factor analysis), path analysis and regression (regression). Another definition mentions structural equation modeling (SEM) is a general and very useful multivariate analysis technique that includes a number of other analytical methods as special cases.

The next definition says that Structural equation modeling (SEM) is a statistical technique to build and to test statistical models that are usually in the form of causal models. SEM is actually a hybrid technique that includes the confirmatory aspects of factor analysis, path analysis and regression.

Advantages of SEM

Slightly different from previous definitions, Structural Equation Modeling (SEM) develops and has a function similar to multiple regression. Although it seems that SEM is becoming a more powerful analytical technique. It ables to consider interaction modeling, non-linearity, correlated independent variables, measurement errors, correlated error terms, multiple latent independent variables. It has many indicators, and one or two latent dependent variables that has also several indicators. Hence, SEM can be stronger than using multiple regression, path analysis, factor analysis, time series analysis, and covariance analysis

Its ability to decipher relationships between variables and to test theoretical credibility (or models). Which uses statistical techniques based on a number of very strict assumptions (Pedhazur, 1982). Three of the assumptions are that the variables the path analysis must be without errors, errors are uncorrelated, and the variables in the model are unidirectional (not including reciprocal models).

In many studies, almost all variables are unobservable variables. Variables such as educational aspirations, anxiety tests, student perceptions are latent behavioral concepts. The use of a single indicator to fully capture something complex in a path analysis is not practical. The nature of these variables in the path analysis requires the involvement of one or several indicators to construct each latent variable.

Path Analysis

Path analysis is an extension analysis technique of the regression model, which is used to test the dependence of a number of variables in a model. This model is generally depicted in a circle image and arrow direction that shows the relationship between the independent variable, mediator variable, and the dependent variable. The model is suitable for testing research hypotheses that show direct and indirect relationships. The causal model formulated by this researcher must be based on strong theoretical considerations.

This path analysis has advantages compared other analysis techniques. It is relatively simple and easy to do. In addition, we can analyze many variables with a relatively smaller number of samples.

The Weakness

As a comparison with SEM Full model, this path analysis only uses the total score of the variables only. While the full model on the SEM is a latent score. Since, it uses total score, that there is no measurement error in one variable. Therefore it is necessary to ensure that the measuring instrument has valid and reliable. Besides, the path analysis also can not see the accuracy index of the model. Therefore between one model with another model can not be compared.

Another weakness of the path analysis is that it does not allow the possibility of a link between residual error values ​​for each endogenous variable. Testing the model with the hypothesis of a joint effect (simultaneous) is rare. It is conceivable, academic experience will not only affect student performance. There is also the possibility of student performance affecting academic experience (eg learning, participating in study groups, accessing academic resources, engaging in class discussions). Thus, the use of path analysis to overcome problems like this is inappropriate.