My research interests include theoretical as well as applied aspects of nonparametric smoothing techniques. In my thesis I did research on different aspects of generalized linear, generalized additive, and single index models. I developed and explored different methods to chose between these models and studied the practical problems arising when implementing these methods (including fast implementation of nonparametric techniques).
With respect to nonparametric smoothing, my main interest concentrates currently on exploring some aspects of ``qualitative'' nonparametric regression. Examples include nonparametric regression under the assumption that one knows that the function is monotone, concave or one has similar qualitative a priori information.
Related to the work on model selection which I did during my doctoral thesis, I recently became interested in two proposals for model selection for ``ordinary'' linear models-namely the ``nonnegative garotte'' (Breiman, 1995, Technometrics) and the ``least absolute shrinkage and selection operator'' (Tibshirani, 1996, JRSS-B). My aim is to adapt these methods to more complicated models and to develop efficient algorithms for fitting such models.