AbstractsMathematics

This thesis deals with the appropriate handling of spatial data in general, and in particular in the framework of economic sciences. An overview of well known methods from the field of spatial statistics and spatial econometrics is given. Furthermore a special class of spatial objects is described, namely objects that are that far apart from all other observations in the dataset, that they are not connected to them anymore. Different treatments of such objects are suggested and their influence on the Moran's I test for spatial autocorrelation is analyzed in more detail. After this theoretical part some adequate spatial methods are applied to the well-known problem of R&D spillovers. The corresponding dataset is not obviously spatial, nevertheless spatial methods can be used. The spatial contiguity matrix is based on an economic distance measure instead of the commonly used geographic distances. Finally, optimal design theory and spatial analysis are combined via a new criterion. This criterion was developed to be able to take a potential spatial dependency of the data points into account. The aim is to find the best design points that show the same spatial dependence structure as the true population. (author's abstract)