Online Program Home
  My Program

Abstract Details

Activity Number: 297 - Advances in Nonparametric Testing
Type: Contributed
Date/Time: Tuesday, August 1, 2017 : 8:30 AM to 10:20 AM
Sponsor: Section on Nonparametric Statistics
Abstract #324329 View Presentation
Title: Multi-Aspect Nonparametric Testing and Ranking on Mixed Distributions with Application to Image Analysis in Neuroanatomy
Author(s): Livio Corain* and Antonella Peruffo and Luigi Salmaso and Enrico Grisan and Jean-Marie Graïc
Companies: University of Padova and University of Padova and University of Padova and University of Padova and University of Padova
Keywords: Multivariate ranking ; Nonparametric combination ; Permutation tests ; Union-Intersection method

When information on membership of statistical units to their reference (sub-)population is unknown, it is advisable referring to mixture distributions. Based on the concept of multivariate stochastic dominance, we propose a nonparametric and permutation-based framework for multi-aspect testing and ranking on mixed distribution populations using image analysis data sets. The proposed methodology provide a flexible and less demanding in terms of underlying assumptions tool to infer on the presence of possible stochastic dominances in both location and scatter aspects. Via a Monte-Carlo simulation study we investigated the proposed method where we proved its validity under different random distributions and type correlations. From the application viewpoint, the proposed methodology can be effective to face comparative problems in image analysis for biomedical research. Finally, we present an application to image analysis in neuroanatomy that through single-cell morphological shape descriptors allows to quantify the cytoarchitecture of a specific brain area and therefore rank the anatomical complexity across different populations considering several factors such as sex, age, species.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

Copyright © American Statistical Association