CB-SEM IMPLEMENTATION IN R AND PYTHON: A REVIEW AND COMPARATIVE STUDY

Abstract

Given the increasing popularity of freely available open-source packages being widely used in academia and corporates, this study explores the implementation of a covariance-based structural equation modelling (CB-SEM) method using R and Python packages. The research model considered for the study is the European Consumer Satisfaction Index (ECSI) which has been frequently used in customer satisfaction and customer loyalty related studies. Moreover, survey-based data gathered from the mobile phone industry as cited in previous research has been used. By simulating different scenarios, namely, missing data, non-normality, and single indicator latent variables, the paper reviews and compares the functionalities provided in ‘lavaan’, an R package, and ‘semopy’, a Python package. This paper provides suggestions to handle these scenarios while implementing the models. It establishes that while both ‘semopy’ and ‘lavaan’ provide comparable results for the measurement model and the structural model, there are a few practical considerations that need attention while using these open-source packages, especially in the case of datasets with missing data and when using models having single indicator latent variables. The developers and researchers using these packages will greatly benefit from this paper and it will enable them to be selective in identifying the right methods and functions for their specific use cases. This study fills the gap in the literature by studying the implementation of CB-SEM using the recently released Python package ‘semopy’ and comparing the results with an established R package ‘lavaan’.