Options
Rakotoasimbola, Eric
Nom
Rakotoasimbola, Eric
Affiliation principale
Fonction
Ancien.ne collaborateur.trice
Identifiants
Résultat de la recherche
Voici les éléments 1 - 1 sur 1
- PublicationAccès libreRigorous testing of causality models in consumer behavior research(Neuchâtel : Université de Neuchâtel - Faculté des sciences économiques - [Institut de l'Enterprise], 2017)
; The hypothetico-deductive method is one of the pillars of scientific research. Within consumer behavior research in marketing, Structural Equation Modeling (SEM) is an increasingly popular statistical approach to deduction, especially given recent software packages that simplify its use. However, facile reliance on the convenience of SEM can create significant pitfalls, particularly regarding its implementation as a statistical tool. To address this issue, this thesis presents three essays on applying SEM more consistently and rigorously. The first essay proposes a decision protocol for using SEM. The second essay assesses the impact of sample size, choice of estimation methods, and degree of nonnormality of variables on fit indices. Finally, the third essay compares the results of two SEM methods, namely confirmatory factorial analysis (CFA) and exploratory structural equation modeling (ESEM), through four involvement measurement models. The general purpose of this study is to assist researchers — especially novices in statistics — in applying SEM, while discouraging those methods that have been shown unsuitable in the literature. The decision protocol suggested in the first essay can also be used by journal reviewers as a benchmark for evaluating SEM-based articles. Specifically, for this essay we reviewed four articles on SEM review practices, and formulated a proposed baseline based on a synthesis of the authors’experiences and recommendations. The second essay emphasizes the importance of choosing the proper fit index when evaluating a model. Mittal and Lee's (1989) causality model of consumer involvement was taken as a model of study. This study demonstrated that the data characteristics (sample size and degree of nonnormality of the variables) and choice of estimation method (maximum likelihood or general least squares methods) affect the fit indices. However, the impacts of each fit index are different. Using the Monte Carlo simulation method, we propose several suggested criteria to assess the Mittal and Lee’s model (1989) for other replication studies. Overall, we recommend the use of Standardized Root Mean Square Residual (SRMR) and Root Mean Square Error of Approximation (RMSEA), but also we suggest Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) under certain specific conditions, depending on the sample size and the choice of estimation methods. Adjusted Goodness-of-FIt statistic (AGFI) and Goodness-of-FIt statistic (GFI) are not recommended. The third essay compares two SEM methods, namely Confirmatory Factor Analysis (CFA) and Exploratory Structural Equation Modeling (ESEM). ESEM is a new approach within SEM, and to our knowledge it has not yet been applied in the field of consumer behavior research in Marketing. We compare these two methods by using the concept of consumer involvement in terms of footwear products. This concept was chosen because the issue of its operationalization has caused much controversy in the literature. After comparing four main models used to measure consumer involvement in CFA and ESEM, we rank their performance according to their psychometric qualities. To perform this analysis, data were gathered by means of a survey among the students of the University of Neuchâtel. The results of the study show that the two methods are not interchangeable, but that ESEM can be a helpful complementary tool to CFA for operationalizing constructs.