The ultrametric covariance model for modelling teachers’ job satisfaction

Multidimensional phenomena are often characterised by nested latent concepts ordered in a hierarchical structure, from the most specific to the most general ones. In this paper, we model a nonnegative data covariance matrix by extending the Ultrametric Correlation Model to covariance matrices. The proposal is a parsimonious model which identifies a partition of variables in a reduced number of groups, and the relationships among them via the ultrametric property. The proposed model is applied to investigate the relationships among the dimensions of the Teachers' Job Satisfaction in Italian secondary schools.