Thomas Augustin

Learn More
Random forests are becoming increasingly popular in many scientific fields because they can cope with "small n large p" problems, complex interactions and even highly correlated predictor variables. Their variable importance measures have recently been suggested as screening tools for, e.g., gene expression studies. However, these variable importance(More)
Simulation studies present an important statistical tool to investigate the performance, properties and adequacy of statistical models in pre-specified situations. One of the most important statistical models in medical research is the proportional hazards model of Cox. In this paper, techniques to generate survival times for simulation studies regarding(More)
The paper studies the extension of one of the basic issues of classical statistics to interval probability It is concerned with the Generalized Neyman Pearson problem i e an alternative testing problem where both hypotheses are described by interval probabi lity First the Huber Strassen theorem and the literature based on it is reviewed Then some results(More)
Dempster-Shafer theory allows to construct belief functions from (precise) basic probability assignments. The present paper extends this idea substantially. By considering sets of basic probability assignments, an appealing constructive approach to general interval probability (general imprecise probabilities) is achieved, which allows for a very flexible(More)
This paper discusses fundamental aspects of inference with imprecise probabilities from the decision theoretic point of view. It is shown why the equivalence of prior risk and posterior loss, well known from classical Bayesian statistics, is no longer valid under imprecise priors. As a consequence, straightforward updating, as suggested by Walley’s(More)
Nonparametric Predictive Inference (NPI) is a general methodology to learn from data in the absence of prior knowledge and without adding unjustified assumptions. This paper develops NPI for multinomial data where the total number of possible categories for the data is known. We present the general upper and lower probabilities and several of their(More)
A new model for learning from multinomial data has recently been developed, giving predictive inferences in the form of lower and upper probabilities for a future observation. Apart from the past observations, no information on the sample space is assumed, so explicitly no assumptions are made on the number of possible categories. In this paper, we briefly(More)
This paper discusses decision making in the practically important situation where only partial prior information on the stochastic behavior of the states of nature expressed by imprecise probabilities (interval probability) is available. For this situation, in literature several optimality criteria have been suggested and investigated theoretically.(More)