Institute for Medical Informatics, Statistics and Documentation, Medical University of Graz



Systematic reviews form an important part of evidence-based medicine (e.g. Cochrane Collaboration). In systematic review work, studies which have been made available to a medical question are seized and evaluated. The quantitative summary of the results of the single studies is accomplished with meta-analytic methods. However, this combination of the results poses a number of statistical and methodical problems.

In the context of different projects we also investigate - for example - how to integrate studies with different study designs into a comprehensive Meta-Analysis. At the same time, it is interesting to demonstrate, which information details must be present from the single studies, in order to be able to combine studies with different study designs. Furthermore, the influence of the study design on the result is examined.

Contact: Andrea Berghold

Randomization Methods

Within clinical studies, a fair comparison of a group of therapies and/or a distortion-free estimation of the effect of the therapy is only possible with an even distribution of known determining factors on the groups of treatments. This even distribution of patients over therapy groups is achieved by randomization.

A row of methods are available for the implementation of the randomization, for example: full randomization, minimization, randomization by means of permutated blocks, urn models, etc. In order to be able to give solid recommendations for the application of these procedures, we conduct simulations for the analysis of the balance behaviour of these methods. Thereby, other influencing factors must also be taken into consideration, for example: stratification or the weighting differences within the therapy groups.

Contact: Petra Ofner-Kopeinig
Team: Andrea Berghold, Petra Ofner-Kopeinig

Regulatory Networks

Medicine has made enormous progress by investigating small subsystems of living organisms. New therapeutic approaches might be found through interventions in the regulation of the complex interplay of single components. However, this requires a deeper unterstanding of the interactions of these components.

Through recently developed high throughput techniques, it is now possible to measure a huge number of parameters of an organism simultaneously, whereby typically a subsystem of the cell metabolism is registered as completely as possible (proteome, genome, transcriptome). Due to the complexity of these regulatory networks, the usual statistical methods are not sufficient. In collaboration with the Institute for Pathology, we are occupied with the development of new statistical procedures for the design of experiments, the description and display of results, as well as the analysis of the network.

Contact: Franz Quehenberger
 Deutsch  :  Contact  :  Imprint  :  Search   » MedUni Graz 
Webmaster, 2011-09-07