About the Course

I will present several new developments dealing with means, correlations, regression and prediction over the past 20 years. One of the driving ideas are attempts to leverage the power of classical statistics applied to new data structures: Networks, graphs, compositional data and non-euclidean spaces. The other basic idea stems from the idea of re-using data to attain efficiency, robustness and finite sample guaranties of optimality. Detailed descriptions of new methods from distance correlation, Frechet means and conformal prediction to likelihood-ratio test without distributional assumptions are complemented by practical examples based on R code.