Contributions to spatial data analysis and Stein's method

The thesis consists of two parts.
The first part is rooted in robust statistics, and more precisely in local outlier detection in spatial multivariate data, and the robustness of tests concerning autocorrelation.
The second part deals with the probabilistic method developed by Charles Stein. In the thesis, Stein's method was linked to first order covariance relations, infinite covariance expansions, and distances between probability distributions.