His research is motivated by complex data, including for longitudinal, clustered and correlated univariate and multivariate responses; classification and improved prediction methods for multiplatform data in but not limited to omics data. He investigates theoretical properties of regularization methods in various asymptotic scenarios and improves these methods by learning from resampling-based stability information. He develops advanced model visualisation methods to enable interactive and dynamic model building, investigates robust selection and estimation methods in regression type models, and devises statistical methods for the analysis of multi-layered and structured data in bioinformatics, microbiome and neuroscience applications.