Probabilistic Structural Equation Modeling Approach to Investigate the Relationships between Passenger Perceived Value, Image, Trust, Satisfaction and Loyalty

Creative Commons License

Karadağ T. , Gölbaşı Şimşek G.

y-BIS 2019 Conference: ISBIS Young Business and Industrial Statisticians Workshop on Recent Advances in Data Science and Business Analytics, İstanbul, Türkiye, 25 - 28 Eylül 2019, ss.77

  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.77


Probabilistic structural equation modeling (PSEM) can be regarded as combination of structural equation modeling (SEM) and Bayesian Network (BN) [1]. One can arrive to the BN phase purely from a structure hypothesized from theory, this is done using classical SEM techniques. As we acknowledge the structure as a SEM, the final BN is then called a probabilistic SEM (PSEM). The other side of the spectrum is a completely data-driven approach, assuming no theory driving the structure. That is an explanatory approach and can be useful in scenarios where data exist, but no extensive theory is available. The resulting BN is then also called an explanatory BN (EBN), rather than a PSEM as the structure is not based on theory as is the case with SEM. If we consider the case where the factors are created according to the theory, but the structural paths are learned using data, the resulting BN is then called a semi-PSEM. The aim of this study is to conduct PSEM to investigate the relationships between the customer satisfaction and customer loyalty considering image, trust, and perceived value in the context of public transportation. In order to explore the relationship between these latent variables, the passenger survey dataset corresponding high speed rail system (HSRS) in Turkey [2] was analyzed. A measurement model of 37 variables for these five latent variables was verified using confirmatory factor analysis (CFA) following explanatory factor analyses (EFA). It was also shown that the 5- factor measurement model was supported by EBN analysis. Latent variable scores obtained by averaging the original 5 point-Likert scale responses of the variables belonging the same latent factor were re-categorized into 5 of classes. This study was also elaborated focusing on the frequency of use by HSRS customers giving attention to the directions of the arc between satisfaction and loyalty as a main concern.       Keywords: Bayesian Network; High Speed Rail System; Loyalty; Probabilistic Structural Equation Modeling; Satisfaction 

 Acknowledgements: This research is supported by the Scientific and Technical Research Council of Turkey (TUBITAK) under the support programme of 3001 (Project No 114K093). References [1] Yoo, K. (2017). Probabilistic SEM: an augmentation to classical Structural Equation Modelling (Doctoral dissertation, University of Pretoria).  [2] Akyıldız Alçura, G., Kuşakcı, Ş., Gölbaşı Şimşek, G., Gürsoy, M. and S. C. Tanrıverdi. 2015. “Impact Score Technique for Analyzing the Service Quality of a High-Speed Rail System.” Transportation Research Record: Journal of the Transportation Research Board 2541: 64-72.