Composite indicators simplify complex phenomena by combining multiple dimensions into a single score that synthesizes the overall latent status, such as gender equality or economic development. These indicators comprise various domains and subdomains, each capturing different aspects. For example, the European Gender Equality Index (GEI) combines data across six domains to provide a score that reflects gender gaps in a country. This paper introduces a multivariate statistical learning approach to measure the gender gap, complementing and enriching composite indicators. In particular, Object-Oriented Bayesian networks are employed. An Italian case-study shows that the model offers insight into territorial disparities impacting gender equality outcomes.