Abstract

The rapid growth of data-driven research across disciplines demands innovative computational tools that can handle complexity, uncertainty, and scale. This session, Advances in Computational Statistics: European Perspectives and Innovations, showcases recent breakthroughs that not only advance statistical theory but also deliver practical solutions for real-world challenges in science, engineering, and social research. From optimizing experimental designs to improving dependence modeling and clustering, these contributions illustrate how computational statistics is shaping the future of data analysis.

The first talk introduces algorithmic tools for experiment optimization, offering strategies to balance competing design criteria in high-dimensional and multi-objective settings. These tools enable more reliable and cost-effective designs in applications ranging from chemical engineering to pharmaceutical research. The second contribution presents a data-driven cross-validation approach for fuzzifier selection in fuzzy clustering, improving clustering accuracy and interpretability in fields such as image analysis, market segmentation, and social science research. The third presentation focuses on tau-informed initialization for maximum likelihood estimation of copulas with discrete margins, a breakthrough for modeling dependence in low-count data, with implications for risk analysis, insurance, and econometrics. Finally, the last session addresses testing independence for incomplete, high-dimensional data, a critical issue for genomics, sensor networks, and other domains where missingness and dimensionality pose significant challenges.

Together, these innovations demonstrate the transformative impact of computational statistics on modern data science. The session will provide clear examples of methods that improve accuracy, efficiency, and decision-making in real-world applications, showing how advanced statistical tools can make complex problems more manageable and solutions more reliable.