The MagIC Seminar Series brings internationally recognized researchers to NOVA IMS to share cutting-edge developments in the key research areas of the MagIC Research Center. These seminars provide an opportunity for researchers, faculty members, and PhD students to engage with leading experts, discuss emerging scientific challenges, and explore new research perspectives and collaboration opportunities.

In this session, Seyed Yaser Samadi, Associate Professor in Computer Science, Mathematics, and Statistics at Southern Illinois University, will present the seminar “Envelope Matrix Autoregression for Economic Time Series”.

The seminar will last approximately 1 hour and 30 minutes, including time for discussion and interaction with participants.

The session is open to all NOVA IMS faculty, researchers, and students.

Matrix-valued time series arise naturally in information management, business analytics, and economic measurement, where complex systems such as trade flows, financial networks, regional indicators, and organizational performance metrics are often observed as matrices changing over time. Standard approaches typically vectorize these data and apply vector autoregressive models, which ignore the inherent row–column matrix structure and result in substantial overparameterization. The matrix autoregressive (MAR) model offers a more structured alternative by preserving the natural matrix form of the data. However, in high-dimensional settings, MAR models may still involve large coefficient matrices and may fail to distinguish relevant dynamic variation from noise.

In this talk, we propose an Envelope Matrix Autoregressive (EMAR) model that integrates envelope methodology to improve estimation and forecasting efficiency for matrix-valued economic and business data. The proposed approach identifies minimal reducing subspaces of covariance matrices, allowing the model to isolate variation that is material to the mean dynamics while removing immaterial information. This leads to meaningful parameter reduction, improved statistical efficiency, and enhanced forecasting performance. We establish the asymptotic properties of the proposed estimators and assess their performance through simulation studies under both normal and non-normal error distributions. Applications to economic and business datasets demonstrate the practical value of the method for data-driven decision-making and advanced analytics.

This event is supported by national funds through FCT,I.P. (Fundação para a Ciência e a Tecnologia,I.P.), under the project - UID/04152/2025 - Centro de Investigação em Gestão de Informação (MagIC)/NOVA IMS (DOI: 10.54499/UID/04152/2025, and by the European Union – NextGenerationEU under the project UID/PRR/04152/2025 (DOI: 10.54499/UID/PRR/04152/2025).

This MagIC Seminar will take place at Executive Room 5, Almada Negreiros College.