Motivation

The goal of data analysis is often to understand, among the increasing number of features, which of these contribute in a statistically significant manner to explaining and predicting a certain phenomenon of interest. Nowadays the growing size of data, and details collected in this data, adds a considerable level of uncertainty in determining which variables must be object of specific analytical focus. Different methods can be used to select the variables to be used within a statistical model but they can sometimes be numerically unstable or not suitable to the problem at hand. Our research aims at delivering approaches that allow to identify the (few) relevant features of a problem in a suitable and flexible manner thereby obtaining significant and highly interpretable results.

Publications

• Guerrier, S., Stebler, Y., Skaloud, J. & Victoria-Feser, M.-P., Wavelet-Variance-Based Estimation for Composite Stochastic Processes, Journal of the American Statistical Association (Theory & Methods), 2013. Link Full text

Conference Proceedings