Upcoming GRBIO seminars
Organizers: Ferran Reverter and Jordi Cortés
Online. If you want to attend, please, contact grbio@grbio.eu
Advances in the analysis of human perceptions within the CUB framework
Questionnaires are widely used across various research fields to assess latent traits such as perceptions, opinions, and attitudes. These traits are typically measured using rating scales like Likert or Semantic Differential scales, generating ordinal data. Analyzing such data presents unique challenges, for which specialized models have been developed. Among these, the CUB (Combination of discrete Uniform and shifted Binomial random variables) class of models stands out as a prominent approach in the literature. CUB models assume that respondents' ratings result from the interplay of two distinct psychological components: feeling, representing a rational response, and uncertainty, reflecting indecision. This talk provides an introduction to CUB models and presents new theoretical contributions to their development. Additionally, an application of these models to analyze the synesthetic experience of museum visitors is presented, showing their practical relevance in studying human perceptions.
Biosketch
Matteo Ventura is completing his PhD in Statistics at the University of Brescia, Italy. His research focuses on both the methodological and applied aspects of Statistical Sciences. His primary interests lie in mixture models and ordinal data, with applications across various fields, including marketing, tourism, culture, and psychology. Additionally, he is interested in graphical models with applications to social sciences and ecology.
Online. If you want to attend, please, contact grbio@grbio.eu
Addressing bias in statistical inference based on epidemiological registry data
Hospital's registry data is a widely used resource in Nordic countries to estimate parameters of interest in the public health and in epidemiology. This data allows the researcher to have an unbiased representation of the population, since it is collected for all the individuals that visit the hospitals in the country, and it is stored in databases that are available for the researchers. Despite being a powerful tool, this data has some drawbacks that must be considered before making the study.
This presentation aims to expose the delayed-entry problem, that arise when hospital's registers are used for estimating the incidence of a disease, and explain the washout periods methodology, that it is usually used to correct (or partially correct) it. We will present a theoretical framework that helps to understand the delayed-entry problem from a formal point of view and a simulation study to analyze this methodology and understand its limitations.
Biosketch
Dídac Gallego is graduated with a bachelor’s degree in mathematics from the Autonomous University of Barcelona in 2022. In the same year, he began a Master’s program in Statistics and Operations Research (UPC-UB), which he completed in June 2024. Currently, he is working at Accenture
Online. If you want to attend, please, contact grbio@grbio.eu
Advances in Time Series Forecasting: The Temporal Fusion Transformer
Time Series Forecasting plays a critical role across many domains, such as supply chain management (e.g., predicting product demand), healthcare (e.g., forecasting hospital admissions to optimize staffing or predicting disease progression), transportation (e.g., forecasting traffic to improve traffic light timings), energy (e.g., anticipating power demand to streamline energy production), or finance (e.g., predicting share prices in stock markets). While classical methods and Machine Learning models have achieved considerable success, they face limitations such as capturing complex temporal dynamics, scaling to high-dimensional data, or providing robust and actionable interpretability. Deep Learning, particularly Transformer-based models, offers a promising alternative capable of addressing these challenges. The Temporal Fusion Transformer is a state-of-the-art Deep Learning model designed for multi-horizon forecasting, and to specifically address many of the shortcomings of prior forecasting approaches. It combines superior performance with interpretability, leveraging its innovative (and customizable) architecture and contextual learning capabilities to address the intricate dynamics of time-dependent data, enabling precise and explainable forecasts, making it a powerful tool for many forecasting applications.
Biosketch
Santiago Llarquet studied Mathematics (2019-2023) and recently completed the MESIO program at UPC. With background in mathematics, statistics, and optimization, he currently works as a Data Scientist. His primary focus is on Time Series Forecasting, where he leverages classical methods, Machine Learning, and Deep Learning models, with a particular interest in Transformer-based models. His work aims to bridge theoretical advances and practical applications, delivering impactful forecasting solutions across multiple domains.
Online. If you want to attend, please, contact grbio@grbio.eu
TO BE ANNOUNCED
Online. If you want to attend, please, contact grbio@grbio.eu
TO BE ANNOUNCED
Online. If you want to attend, please, contact grbio@grbio.eu
TO BE ANNOUNCED
Online. If you want to attend, please, contact grbio@grbio.eu
TO BE ANNOUNCED
Online. If you want to attend, please, contact grbio@grbio.eu
TO BE ANNOUNCED
Share: