Conferencia plenaria en Estadística / Plenary lecture in Statistics
|
Simon N.
Wood School of Mathematics University of Bristol United Kingdom |
Smooth Regression Beyond the Exponential Family
Regression models specified in terms of smooth functions of predictors are widely applicable in fields as diverse as fisheries, epidemiology and ecology. Their advantage is that they possess a flexibility that mitigates against model mis-specification. But flexibility brings challenges, both computational and statistical: stable computation can become a difficult task and we also need methods for objectively choosing the appropriate degree of smoothness of the model components. Within the exponential family of distributions, quite reliable methods are now available addressing both these challenges. But can the same be achieved in more general settings outside the exponential family? This plenary lecture will review some of the options, taking examples from location scale modelling, smooth survival analysis and quantile regression.
Conferencia plenaria en Investigación Operativa (Conferencia Sixto Ríos) / Plenary lecture in Operations Research (Sixto Ríos Lecture)
|
Ignacio García
Jurado Departamento de Matemáticas Universidade da Coruña |
Algunos Modelos de Investigación Co-Operativa
Podríamos definir la investigación co-operativa como la aplicación de métodos matemáticos para el apoyo a la toma de decisiones en situaciones en las que varios agentes cooperan en la puesta en marcha y funcionamiento de un sistema complejo.
La resolución de un problema de investigación co-operativa conlleva tres fases:
- identificar un modelo matemático que describa el problema,
- encontrar una solución óptima para el modelo en el caso de que todos los agentes cooperen (usando técnicas de optimización),
- repartir entre los agentes los beneficios que resultan de su cooperación (usando técnicas de juegos cooperativos).
En esta conferencia se presentará una breve introducción a la teoría de juegos cooperativos y se analizarán algunos modelos de investigación co-operativa en el contexto de los problemas de optimización de flujo en redes, de los problemas de inventario y de los problemas de planificación de proyectos.
Conferencia plenaria en Estadística Pública / Plenary lecture in Official Statistics
|
Eduardo Barredo Eurostat Dir B - Methodology, Corporate Statistical and IT Services |
Skills for the new generation of statisticians
At the core of the modernisation of official statistics will be the capability to incorporate new data sources and to take benefit from disruptive technologies, for example smart meters, web technologies and user experience platforms. These new capabilities require new types of skills and competences that were not specifically found in the traditional official statistician skill sets. In this paper, we analyse the competence profile of an official statistician with a particular focus on new data science competences.
There are numerous examples of new data sources that have potential in this sense: administrative records and registers, as well as large digital data sources, such as road sensors, scanner data or Internet-based data. These large digital data sources are known best as big data. Potential new technologies, on the other hand, range from web-scraping algorithms to linked data opportunities and from multi-mode data collection to combining survey and administrative data. The list can be continued even further.
By embracing new data sources and technologies, National Statistical Institutes (NSIs) can produce faster, more accurate statistics and more comprehensive indicator sets adapted to understanding of increasingly complex, rapidly changing and global phenomena. In short, new data sources are a way to better meet users’ needs. This must be done in a way where we do not jeopardise the recognised robustness and quality of official statistics in order to strengthen our competitive asset in the rapidly changing information ecosystem.
One key factor in meeting these challenges is the development and building of the necessary skills and competences. In addition, statistical organisations will have to create favourable conditions for new production methods and using data science skills with success. This means, for example, 1) establishing an innovative culture, where experimental activities are commonplace, 2) building and maintaining collaborative and multidisciplinary data science teams, 3) recruiting individual data scientists and using long-term personnel planning, 4) training and supporting personnel in identifying themselves with new competence requirements, and 5) aspiring to management and leadership practices, which make these changes possible.
