Phd in SCHOOL OF STATISTICAL SCIENCES

email: donatella.vicari@uniroma1.it
phone: (+39) 06 49690210 (int. 30210)
building: CU002 - Statistica
room: 12

Full Professor in Statistics

The large majority of my research is mainly focused on the analysis and development of methodologies in the general context of classification and dimension reduction, with reference to both theoretical and computational issues.
In all papers, "simultaneous models" are proposed which jointly optimize a single loss function (for example concerning classification and/or dimension reduction and/or regression), with a substantial improvement in terms of performance over the application of sequential procedures which lead to sub-optimal solutions for the general problem.
In several papers the presence of possible external variables has been also considered in a confirmative approach.
In particular, the topics of most publications, also in collaboration with Italian and foreign scholars, cover different types of data:
- Three-way proximity data
Most of the proposed methodologies concern the analysis of three-way proximity data where pairwise similarities between objects are observed in different occasions/subjects. Models are proposed for clustering the objects in either a finite mixture or a least-squares framework. The heterogeneity is accounted for by deriving classes of similar occasions/subjects using different models.
- Asymmetric proximity data
A recent chapter of research is entirely devoted to the development of methods and tools for clustering asymmetric proximity data (i.e., for example confusion data, flows, brand switching data). In such papers the novelty is the proposal of clustering models where the directions of the exchanges have been accounted for in addition to their average amounts.
A monograph on this topic coauthored with prof. G. Bove (University of Roma Tre, Italy) and A. Okada (Tama University, Japan) has been published in 2021 (Springer Nature).