|
General Theory and Methods |
Departament Informàtica i Matemàtica Aplicada, Universitat de Girona, Campus Montilivi, Edifici P-IV, E-17071, Girona, Spain
(e-mail: josepantoni.martin{at}udg.es)
It is very important to realize that the well-known zeros problem in compositional data is inherent in the nature of the data rather than in the log-ratio methodology. In a strict sense, any null value is informative itself and needs specific treatment before a multivariate method is applied. In the rounded zero case specific techniques of missing data should be applied as a previous step, taking into account that any of these techniques must respect the compositional nature of the data. In practice, when an imputation method is applied, then it is necessary to make a sensitivity analysis of the results from multivariate analysis. These methodological aspects are applied and illustrated using compositional data from geological samples.
This article has been cited by other articles:
![]() |
V. Pawlowsky-Glahn and J. J. Egozcue Compositional data and their analysis: an introduction Geological Society, London, Special Publications, 2006; 264: 1 - 10. [Abstract] [PDF] |
||||