FCT

R&D Institutions

Resultado da avaliação 2007 na área de Matemática

Unidade de I&D

Centro de Estatística e Aplicações da Universidade de Lisboa [MATH-LVT-Lisboa-6] visitada em 20/02/2008

Classificação: Very Good

Comentários do painel de avaliação
Sobre a unidade
The Statistics group at the University of Lisbon is the largest and oldest in Portugal, with a particular historical strength in extreme value theory. During the period 2003-2006 they have continued to develop that interest and are now clearly one of the leading research groups in the world for that subject. At the same time a second also very strong group has developed in applications to environmental modeling and life sciences. However, two other groups are less impressive. The overall ranking of the center is VERY GOOD and clearly reflects the disparity among these different groups.
Sobre os grupos de investigação
Applied Multivariate Data Analysis and Modelling [RG-MATH-LVT-Lisboa-6-1519]
This is a respectable research group with good contacts with social and medical researchers. However the mathematical level of the research does not seem to be in the same class as Group 3, involving less original methodology and not appearing in the top statistics journals. Indeed there was some doubt whether the group really belongs within Mathematics though given the lack of an obvious alternative, continued membership of CEAUL probably is in the best interests of the group themselves. There is evidence of international collaboration but not really of international recognition of their work. They have several externally funded industrial research projects, though by the group's own admission, the potential of these projects to lead to high-quality publications is low. The group has an excellent record of producing MSc students, as well as some PhDs. For the future, it is difficult to give concrete recommendations since the direction of the group seems set, but the research ranking would certainly be improved if there were at least a small number of publications in major statistics journals.
Order Statistics, Extremes and Applications [RG-MATH-LVT-Lisboa-6-948]
This group does theoretical research with a high level of mathematical sophistication. They have a high level of visibility within the international community of extreme value theory, and have contributed extensively to the journal Extremes and to the series of international conferences on extreme value theory that have been organized in recent years. However, some of the work is rather narrow and is not recognized within the broader statistics community outside extreme value theory.

During the period of the review, they have been considerably strengthened by the long-term participation of Professor Laurens de Haan, who spends about 75% of his time in Lisbon. De Haan and a co-author who is newly joining this center published a book with Springer in 2006 which is already widely cited as one of the leading reference books on extreme value theory. De Haan also continues his own research on multivariate extremes and max-stable random processes, which has a close connection with spatial statistics and much potential for application. The application side is somewhat limited at the moment, but a recent joint paper on rainfall extremes (the co-authors include a leading Dutch meteorologist) shows the potential for this work to reach a wider scientific audience.

The other focus of the group, on estimation of the tail index and related problems, is narrower, and while there have been real and deep mathematical advances, this may be less suitable as a subject in which to train young researchers. Applications in other areas, such as quality control, are also being developed.

Overall, the classification of excellent reflects the theoretical quality of the work though this is heavily influenced by the personal contributions of de Haan. For the future, it is appropriate to continue to work on theoretical research but a stronger effort could be made towards applications. Production of both PhD and MSc graduates is excellent. The group has two programmatic funds in addition to their base funding.
Probability, Modelling and Data Analysis [RG-MATH-LVT-Lisboa-6-949]
The focus of this group is far from clear. The written report (and the unit coordinator’s introduction on the morning of the site visit) emphasized such topics as stopped sums and extremes, and small-sample asymptotic, but these topics did not appear in the list of publications, nor were they addressed by the presentation from the group coordinator. As a result, the panel was left in confusion about the group's true objectives. As far as the panel can tell, the main emphasis of the group is on applied statistics, apart from some theoretical papers on the topic of Group 1. There is also work in the biology/ecology area, such as collaboration with a prominent statistical ecologist in the U.K. This also appears to be the strongest evidence of international recognition of the group's work. The research productivity of the group is good, though generally not in top journals, and the production of PhD and MSc theses (e.g. 4 PhDs during the review period) is respectable. Overall, there is clearly evidence of research competence here, but the group needs a sharper focus if they are really to achieve their objectives.
Statistical Modelling in Environmental and Life Sciences [RG-MATH-LVT-Lisboa-6-951]
This appears to be the strongest group overall within this center. During the review period they have continued to diversify their range of applications, to the point where it is now very broad. However the statistical content is not applied statistics in a narrow sense, but includes serious statistical modeling and computation involving original programming. As a result, the group's best papers include some in highly visible journals. This may be the only group in Portugal that is making serious advances in Bayesian statistics but that is important for the visibility of the country's research. They have excellent international contacts including one member’s vice-chairmanship of the SPRUCE series of workshops on environmental statistics. Production of PhD and MSc students has also been excellent. They have one programmatic funding grant in addition to their core funding.