Links to the RTG's Research Projects:
Ongoing Research Projects
Finished Research Proects
Scaling problems in applied statistics have been well known for more than 70 years but still remain a great challenge to this day. No single preeminent statistical approach exists, but a variety of problem-dependent approaches have been recognized, including geostatistical methods, Bayesian hierarchical models, and other multiscale processes. Kernel based methods are the link between the three areas of application investigated within the RTG (economy, ecology, and genetics). They range from geostatistics to geoadditive mixed models. Furthermore, stochastic geometry and Monte Carlo techniques form subordinate bundles of methods.
The functioning of markets as a means of enhancing the welfare of individuals and other economic agents crucially relies on the transmission of prices, technologies, and knowledge throughout the economy. This requires that information, particularly on economic opportunities and scarcities, can flow quickly and without friction between economic agents from the micro to the macro level, and vice versa. The complex mechanisms of transmission along the various stages of the value chain, which stongly depend on scale, interlock in space and time. Statistical modelling of market integration, productivity decomposition, and the impact of policies on livelihoods is necessary to understand these mechanisms better and to generate improved policy recommendations.
In ecology, the fact that a single plant individual can hardly be sustainably managed leads to basic questions such as how to define a forest, or how to measure basic properties such as the edge of a forest. Further, the role of false homogeneity assumptions for the prediction of forest dynamics will be investigated. In agro-ecosystems, we study how ecological interactions in the crop landscape change with scale, and how this influences agricultural productivity. Specifically, we will test the spillover theory and examine barrier effects of landscape elements to organism dispersal. In addition, we will study the distribution and ecological interactions of harmful bacteria on leaf surfaces at different scales to advance biological control strategies for an efficient crop management.
Extensive genomic data are a common challenge both in human genetics and in farm animal breeding. Approaches are needed to deal with whole genome data for the association between genotypes and phenotypes. The steps from coarse (6 · 10^4 markers) to dense (10^6 markers) to whole genome sequences (3 · 10^9 basepairs) pose a major scaling problem. The goals reach from corrections for genetic distances between individuals/populations, to identification of novel disease pathways and to prediction of phenotypes based on aggregate genotypes using, e.g., new, semi-parametric approaches in prediction. Coalescent-based approaches will be developed and used to detect and better understand selection and to assess the impact of various scale issues on the outcome.