DUGi

Statistical Modelling of Compositional Problems Involving Finite Probability Distributions

http://dugi.udg.edu/item/http:@@@@hdl.handle.net@@2072@@273652

Finite probability distributions and compositional data are mathematically similar, consisting ofD-dimensional positive vectors with sum 1. Despite this similarity the meaningful forms of analysisin these different areas may require substantially different concepts and statistical modelling. Thispaper highlights these differences, but also poses the question of how such differences may contributeto understanding in the different areas. At CoDa workshops we have become so accustomed to, evenobsessed with, modelling all compositional data problems within a simplex sample space togetherwith its algebraic-geometric Hilbert space structure. The context of this Hilbert sample space iscertainly often relevant to the formulation of a number of compositional data problems, but itsmathematical elegance should not override appropriate meaningful statistical modelling to resolvethe real compositional problem. In this paper I illustrate some relevant modelling by considerationof how a variety of persons differ in their ability to perform inferential tasks particularly in theprocess of differential diagnosis

Universitat de Girona. Departament d’Informàtica i Matemàtica Aplicada