Motivation

When performing statistical modelling and inference, the analysis often relies on the assumption that the observed data is an exact realization of a certain probabilistic model. However, in the majority of cases, this assumption is not respected since there are different sources of contamination in various applications and, even if extremely small, these can severely bias results and interpretations. For this purpose, our research aims to propose robust solutions to the various approaches that we develop, thereby greatly limiting the impact that these data contaminations can have in different (complex) settings.

Publications

• Guerrier, S., Molinari, R. and Victoria-Feser, M.-P. Estimation of Time Series via Robust Wavelet Variance, Austrian Journal of Statistics, 2014. Link Full text

Conference Proceedings