Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Compositional latent variables, logratios and logcontrasts: underlying concepts in multivariate methods for compositional data
Josep-Antoni Martin-Fernandez

Last modified: 2013-06-14

Abstract


Compositional data (CoDa) require an appropriate statistical analysis. Methods based on log-ratio approach give a consistent framework to analyse CoDa. When the manifest variables are compositional, the latent variable model should be formulated according to the geometry of the simplex. This geometry includes the log-contrast, a simple way to express a set of log-ratios in a linear form. Basic concepts and properties of log-ratios and log-contrast are revisited. Implications of these elements for the latent variable model are explored.

Full Text: PDF