Gaussian graphical models: contributions for exploratory data analysis in organisational behaviour[Les modèles graphiques gaussiens : quels apports pour l’analyse exploratoire des données en comportement organisationnel ?]

Methodological issues arising from access to large data sources are now affecting domains of research that were previously not very concerned, such as organisational behaviour. The discussion on methods for taking advantage of the possibilities offered by large amounts of secondary data is relatively recent. Management scholars, traditionally accustomed to working with small samples in a deductive framework, face a real methodological challenge when they seek to benefit from secondary data through a data-driven approach. One possible approach to meet this challenge is the use of Gaussian graphical models (GGMs), which allow for the visualisation and analysis of relationships between a set of Gaussian variables. The application of this approach to psychology has led to the development of a very active line of research, known as Network Psychometrics, which is renewing the study of attitude measurement scales by relying on parsimonious graphs. The aim of this article is to illustrate the potential added value of this approach in the field of organisational behaviour. We will show that GGMs can offer a complementary point of view when it comes to analysing systems of interactions between variables and we will discuss how they can be articulated with confirmatory approaches using structural equation methods, more commonly used for this type of analysis. The challenges of this articulation will be illustrated by exploring the French version of a recent measure of workplace commitment.