Pesquisadora da Nottingham Trent University ministra seminário no LNCC

No dia 29 de julho, às 11h, o Laboratório Nacional de Computação Científica (LNCC) realizará o seminário:“From Generalised Poisson Models to Scientific Discovery: Advances in Bayesian Modelling, Spectrophotometry and X-Ray Fluorescence Imaging”. A palestra será ministrada em inglês por Golnaz Shahtahmassebi, da Nottingham Trent University. O evento acontece presencialmente no auditório A do Laboratório. 

Resumo

The Poisson distribution remains one of the fundamental probabilistic models for count data and rare events across science, engineering and finance. However, many real-world datasets violate its underlying assumptions through overdispersion, underdispersion, non-homogeneous processes, or integer-valued outcomes extending beyond the non-negative domain. These limitations have motivated the development of more flexible statistical models capable of accurately representing complex count processes.

This presentation traces the evolution of our research on Poisson-based modelling over the past decade, beginning with the development of the Generalised Poisson Difference (GPD) distribution and its Bayesian framework for modelling integer-valued data. The methodology was successfully applied to ultra-high-frequency financial data and football score differences, demonstrating improved flexibility over classical Poisson difference models while naturally accommodating excess zeros and dispersion effects.

Building upon these theoretical foundations, the talk explores emerging interdisciplinary applications. The first examines photon-count statistics in modern spectrophotometry, where our recent work challenges the conventional assumption that spectrophotometers are Poisson-limited and demonstrates how statistical modelling provides new insight into instrument noise and measurement uncertainty. The second presents ongoing research applying Generalised Poisson models to energy-channel counts in X-ray fluorescence (XRF) imaging, where improved modelling of count dispersion enhances elemental analysis, clustering performance and materials characterisation in cultural heritage and analytical science.

Together, these studies illustrate how advances in statistical methodology can translate into practical solutions across diverse scientific disciplines, highlighting the continuing importance of flexible count-data models for modern data analysis and imaging technologies.

Breve Currículo

A Professora Associada Golnaz Shahtahmassebi é Estatística Aplicada na Nottingham Trent University (Reino Unido) e Professora Adjunta na Walter Sisulu University (África do Sul). Sua pesquisa concentra-se em estatística bayesiana, inteligência artificial, processamento de imagens, sensoriamento remoto e modelagem computacional, com aplicações que abrangem ciência ambiental, patrimônio cultural, medicina e saúde pública. Ela lidera colaborações internacionais e interdisciplinares na Europa, América do Sul, África e Ásia, desenvolvendo métodos orientados por dados que traduzem informações complexas em tomadas de decisão baseadas em evidências. Seu trabalho atual inclui inteligência artificial aplicada à ciência do patrimônio, clima e saúde da mulher, além de modelagem bayesiana dinâmica para sistemas populacionais e ambientais.

Serviço de Comunicação Institucional
secin@lncc.br