A doctoral student in management spent three months self-studying Structural Equation Modeling from SPSS manuals and YouTube videos. After finally grasping the basics, they realized that the software they were learning was not optimal for the model they designed and had to start over with SmartPLS. This is not an uncommon story. SEM is one of the most powerful yet mysterious analysis techniques for many Indonesian researchers, and choosing the wrong approach at the outset can waste months of time.
Structural Equation Modeling with Partial Least Squares (SEM PLS) is a multivariate analysis approach that allows researchers to simultaneously test complex causal relationships between latent variables (constructs that cannot be directly measured) and their indicators. Compared to covariance-based SEM (CB-SEM), PLS is more flexible in terms of sample size, data distribution, and model complexity, making it popular in business, management, and social science research where ideal data conditions are often not met.
Building a Scientific Model: From Variables to Constructs
The most often overlooked stage in SEM PLS is mapping variables and building solid constructs. Many researchers rush into the analysis stage without ensuring that their constructs are clearly defined, indicators are selected based on theory, and the direction of measurement (reflective vs formative) is accurately determined. Errors at this stage cannot be corrected by any software; garbage in, garbage out fully applies in SEM.
The combination of bivariate (two-variable) and multivariate (multiple-variable) analysis within the SEM framework allows researchers to build a more realistic model of complex social phenomena. A customer satisfaction model, for example, can simultaneously test the effects of service quality, price, and brand image on loyalty, while considering mediating and moderating variables. This is something that ordinary regression analysis cannot do.
R Studio as a SEM PLS Platform: Why It’s Worth Choosing
Although SmartPLS is more popular among Indonesian researchers, R Studio offers significant advantages for SEM PLS: it is free without a license, more flexible for customizing analyses, and its results are easier to reproduce in the context of scientific publication. Packages like seminr and cSEM enable complete SEM PLS implementation, from model estimation, testing reliability and validity (AVE, CR, HTMT), to bootstrapping for statistical inference.
Model visualization in R is also much richer: publication-ready path diagrams, plots of convergent validity, and goodness-of-fit indices can all be generated with documented and replicable code. For researchers aiming to publish in reputable international journals, transparency and reproducibility of analysis are increasingly valued by reviewers.
AI as a Research Collaborator: Not a Shortcut, But an Accelerator
Integrating AI into the SEM PLS process is not about letting AI analyze the data. AI acts as an accelerator at several critical stages: helping researchers identify relevant indicators based on literature reviews, conducting initial construct validation by assessing the consistency of conceptual definitions, and assisting in interpreting statistical output in more understandable language. The proper use of AI in research does not replace researchers' thinking; it accelerates processes that previously required lengthy iterations.
With the combination of SEM PLS, R Studio, and AI, Indonesian researchers now have access to an analysis toolkit that was previously only available in the research labs of the world's top universities. The barrier is no longer access to software or knowledge; the barrier is the willingness to learn how to integrate all these tools into an efficient research workflow. For those who succeed, the quality of research produced can surpass the limitations of resources that have long been considered obstacles.
References:
- Hair, J.F. et al. (2022). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (3rd ed.). SAGE Publications.
- Rademaker, M.E. & Schuberth, F. (2020). cSEM: Composite-Based Structural Equation Modeling. R package → CRAN.
- Ray, S. et al. (2021). seminr: Building and Estimating Structural Equation Models. R package → CRAN.