Tentative Abstract:

I work with methodogy for generalised linear models where the number $p$ of covariates is larger than the number $n$ of individuals. Methods more general than e.g. those of ridging emerge when priors are used that correspond to mixtures of lower-dimensional structures. These lead to strategies for finding say the $k$ most promising covariates, for $k$ not exceeding a user-defined threshold number $k_0$. The methods are applied to survival data in combination with gene-expression data.