Philipp’s research broadly revolves around improving estimation and prediction for a broad class of statistical models.
In particular, Philipp's recent focus lies on maximum penalized likelihood frameworks for models where the original maximum likelihood estimator does not exist under certain data configurations, or where signal recovery and inferential performance break down in high-dimensional settings.
Further, Philipp is working on a novel approach towards first-order unbiased predictions in statistical and machine learning models. Amongst others, this methodology encompasses the large class of predictive models where training is conducted through optimisation of some loss function (e.g. maximum likelihood estimation or Neural Networks).
Prior to joining ÐÓ°ÉÂÛ̳ as a Fellow, Philipp obtained a PhD in Statistics at the University of Warwick and has a background in Economics (MPhil Cambridge University, BA University of St. Gallen) and Mathematics (MSc ÐÓ°ÉÂÛ̳).