Events and Meetings of Italian Statistical Society, Statistics and Demography: the Legacy of Corrado Gini

Font Size: 
Clustering and Classification methods for an experimental study on prion diseases.
Giorgia Rocco, Luca Tardella, Michele Di Bari, Romolo Nonno, Maria Puopolo

Last modified: 2015-09-05

Abstract


We report on a hybrid approach to analyze a dataset derived from an experimental study on prion diseases conducted at the Istituto Superiore di Sanita`. The data comes from inoculating different strains (inocula) of the disease to bank voles. The aim of the research is to understand at what extent some phenotypic outcomes such as survival times and profiles of brain lesions are able to detect the underlying heterogeneous multi-level origin of the data. We use first an ensemble of hierarchical clustering through the Gower index, a general coefficient that includes similarity for different metrics in the dataset (quantitative and ordinal data). We have verified the ability of the proposed approach to match some preliminary knowledge on the underlying group structure with some possible hint at detecting a slightly finer structure. We then consider alternative classifiers with the aim of validating alternative clustering structures and predict whether a new observation belongs to a group or another.


Full Text: PDF