Choosing an adequate statistical model for an applied problem is the ﬁrst and often most difficult step in any empirical data analysis. In many practical applications, the statistical behaviour of spatial or temporal data depends particularly on the chosen scale, and the data are ﬁtted to different, sometimes contradicting models across scales. The aim of this Research Training Group is to develop solutions to problems of scale that arise in current areas of both scientiﬁc and public interest. This comprises the identiﬁcation of adequate methods and their modiﬁcation as well as the development of new statistical models and methodology.
Three areas of application will be considered:
- environmental sciences,
Methodologically, they are embraced by two branches of statistical methods based on covariance functions (i.e. reproducing kernel Hilbert spaces): mixed models and geostatistics. As regards impact on applied research, they strongly overlap in questions concerning welfare and food, including poverty, biodiversity (including degradation), and biological control. Building on interdisciplinary cooperation and the associated synergistic effects, we expect new impulses for research and more advanced solutions to complex practical problems.
For details on research projects, see Research.
PhD students will become aware of the omnipresence of scaling problems, and will be introduced to a spectrum of approaches ranging from parametric to non-parametric methods. Furthermore, they will beneﬁt from interdisciplinary work and a study environment in which the instruction, supervision, and support of PhD students across disciplines and faculties is standard.