Proposal Description

Modern statistical analysis increasingly involves data structures that go beyond classical single-source, single-valued and independently observed units. Data may be heterogeneous, multi-view, high-dimensional, spatially and temporally dependent, functional or observed over fragmented time windows. These settings require statistical and computational methods able to account for dependence, heterogeneity, uncertainty, complex representations and interpretability. This IASC IPS will present recent advances in complex data analysis, with emphasis on both methodological development and relevant applications. The proposed contributions cover several complementary perspectives on modern complex data.

One contribution focuses on the integration of multi-source geophysical signals for the identification of predictive earthquake regimes. A second contribution addresses high-dimensional and mixed complex data through sparse weighted consensus clustering for feature selection. This work is motivated by the need to improve variable-importance assessment in Random Forests when predictors are highly correlated and of mixed type. A further contribution develops a Bayesian block time-series framework for count data observed over disjoint temporal windows. The method represents irregular time series as sequences of observed blocks with local dynamics while modelling dependence across blocks, providing a flexible alternative to both complete independence and global stationarity assumptions Finally, the session will include recent developments in Bayesian clustering of functional data.

Together, these contributions illustrate how complex data analysis can support clustering, classification, feature selection, dynamic modelling and Bayesian inference across a wide range of domains, including earth sciences, astronomy, economics and other data-intensive scientific fields. The session will highlight the role of IASC in promoting methodological innovation for modern data structures and scientifically relevant applications.