Research Area
Basic Research
Start Date
2009-01-01
End date
2012-12-31
Status
Finished
Project leaders
de Uña-Álvarez, Jacobo
Members
de Uña-Álvarez, Jacobo; Pardo Fernández, Juan Carlos; Iglesias Pérez, María del Carmen; Cotos Yáñez, Tomás R.; Álvarez Díaz, Marcos; Costa da Conceiçao Amorim, Ana Paula; Liang, Han-Ying; Gonçalves de Macedo Moreira, Carla; Veraverbeke, Noel; Rodríguez Girondo, Mar.
Description
Nonparametric and semiparametric methods have become a flexible statistical tool in exploratory data analysis and inference, including: curve estimation, goodness-of-fit testing, regression analysis, resampling methods, multivariate statistics, time series forecasting, or spatial data analysis. This project includes a number of methodological advances in this area. The motivation comes from specific real life problems for which proper statistical tools are missing. Explicitly, one of our main focus will be solving nonstandard issues in the statistical analysis of survival data coming from medical sciences (similar examples being found in the econometric analysis of duration or transition data, and in reliability studies). Non-Markov multi-state models, multivariate survival analysis, models for bivariate censoring, singly or doubly truncated data, censored and/or truncated dependent data, dimension reduction methods, location-scale models, additive censored regression, models for dependent censoring, or presmoothing methods for informative censoring, are some of the specific, modern problems we address. Also, in the field of engineering and environmental sciences, one often has to face large data sets showing a complex pattern of spatial (or time-space) dependence. Our plan is concerned with the development and application of new flexible tools in this context, including nonparametric variogram estimation, kriging, expert systems design, or classification methods, under the useful view of fuzzy sets theory. As a third example, we consider time series forecasting in finance, tourism and the environment, via nonlinear and nonparametric methods. Some techniques which have been deeply investigated during the recent years and new advances will be used to this end. We include as a task the development of new software to implement the new proposed methods. To this regard, this project promotes the using of R as a free programming and data analysis language, so any applied scientist will have access to the routines we will generate