Path Analysis with Composite Variables

Area of Science:

• Statistics
• Quantitative Psychology
• Econometrics

Background:

• Traditional path models often rely on latent variables, which can be complex to implement.
• Composite variables offer a practical alternative, simplifying model fitting.
• Soft modeling approaches, like Partial Least Squares, are gaining traction for their flexibility.

Purpose of the Study:

• To introduce and evaluate six methods for fitting path models using weighted composite variables.
• To connect these composite variable methods with Partial Least Squares (PLS) soft modeling.
• To establish criteria for comparing the performance of these path modeling techniques.

Main Methods:

• Introduction of six distinct methods for path model estimation with composite variables.
• Implementation details for five methods using conventional statistical software.
• Application of Partial Least Squares (PLS) as a framework for soft modeling.

Main Results:

• Five of the six introduced methods are easily implementable with standard software.
• Development of specific criteria for evaluating and comparing the performance of the fitting methods.
• Comparative analysis and evaluative remarks on the devised methods.

Conclusions:

• Composite variable approaches provide accessible alternatives for path modeling.
• The proposed criteria facilitate the selection of appropriate methods for specific research contexts.
• Further research can build upon these methods for advanced soft modeling applications.