PLSsemEngine: A transparent R implementation for composite-based Mode A estimation of reflective models

Partial least squares structural equation modelling (PLS-SEM) is widely applied in the social and management sciences due to its flexibility and limited distributional requirements. However, many existing implementations incorporate extensive automation and implicit decision rules that may reduce computational transparency and obscure analytical control. This paper presents a transparent and modular PLSsemEngine implemented in base R, specifically restricted to Mode A composite estimation. Acknowledging the fundamental methodological distinction between true common factors and emergent composites, this engine is deliberately designed to provide the traditional composite-based approximation of reflective measurement models. The software provides a fully inspectable workflow that structurally separates estimation, resampling, and predictive evaluation into distinct modular functions. Core functionalities include iterative latent score estimation, measurement model assessment, structural path estimation with bootstrap inference, predictive evaluation using k-fold cross-validation (PLSpredict) and selected diagnostic procedures. The implementation deliberately avoids automatic model re-specification, threshold-based classifications, and hidden heuristics. Results are returned as structured, publication-ready tables within a reproducible S3 object. An illustrative example demonstrates the typical workflow and output structure. The engine is intended for researchers and educators seeking a lightweight, reproducible, and methodologically transparent PLS-SEM implementation in R.