Document Actions

You are here: Home Online Magazine connect & create 30 years of Data Analysis and …

30 years of Data Analysis and Modeling in Freiburg

The interdisciplinary seminar that became an academic center of the University of Freiburg

Freiburg, Oct 27, 2017

30 years of Data Analysis and Modeling in Freiburg

Photo:Sandra Meyndt

For 30 years now, academics from different disciplines have met up at the University of Freiburg to exchange ideas about methodical issues of quantitative research – looking at how to obtain more information from data and describe complex systems better with mathematical models. In the beginning they met for a seminar, and since 1994 at the Freiburg Center for Data Analysis and Modeling. The anniversary celebrations take place on Friday, 3rd November 2017.

The Freiburg Center for Data Analysis and Modeling is based on the premises of the Institute of Mathematics at the University of Freiburg.
Photo: Sandra Meyndt

The basis for Big Data – the management and analysis of large quantities of data – is mathematical statistics and stochastics, which deal with chance and probability. Using them demands care, especially in science, “If researchers let program packages loose on their data in order to analyze them using statistical methods, they need to know what the programs can and can’t do. There are also many examples of complex systems that can only be reasonably described with stochastic models – from the flow properties of certain materials to events on the financial market,” says Josef Honerkamp. Thirty years ago, the University of Freiburg professor of physics – now an emeritus – brought colleagues from mathematics, physics, chemistry, medicine, economics and forest sciences together to look at methodical questions like these in a joint seminar. This was the birth of the Freiburg Center for Data Analysis and Modeling (FDM), which eventually became a full academic center of the University of Freiburg in 1994.


VIDEO: Josef Honerkamp on data analysis and modeling

In the early years, this interdisciplinary approach was new, and the advantages for all involved quickly became apparent: mathematics was able to refine its methods thanks to new concepts from a variety of areas, other disciplines made advances because they could draw more information from their data through using statistics and stochastics and describe complex systems better with mathematical models. There have been many more developments since, says stochastics expert Prof. Dr. Peter Pfaffelhuber, director of the FDM, “Since we established our center we have experienced quantification in many areas of application, and now the academic landscape can’t do without interdisciplinary approaches.” In recent decades especially, the FDM has developed dynamically, not least thanks to the appointment of academics working in statistics in different faculties.

Five subject areas in focus

Currently, the work of the FDM focuses on five subject areas: Clinical Epidemiology looks at the spread and progression of diseases and their causal factors in the population; Financial Market Research is dedicated to modeling economic opportunities and risks; Population Biology researches the structure and development of animal and vegetable reproductive communities, and System Biology works on dynamic modeling of cellular processes. Finally, the Statistics of Stochastic Processes researches the mathematical principles of the applications that are established in the center. “It is especially important to maintain a joint methodical core precisely because the areas of application are so varied,” stresses Pfaffelhuber. “The original vision of a space for exchange of ideas on the subjects of data analysis and modeling that would attract a lot of academics has proven itself over the past 30 years.”

The ’flu medicine Tamiflu has a positive effect – but it does not increases chances of survival.
Photo: naihoet/Fotolia

Case study 1: Clinical Epidemiology – how the ’flu remedy Tamiflu works

Many states stockpile the influenza medicine Tamiflu in preparation for a potential pandemic – but it is debatable how well it works. An observational study commissioned by the pharmaceuticals industry compared the disease progression of patients, some of whom were administered Tamiflu, and reached the conclusion that the medicine increases chances of survival. In order to review the study, Prof. Dr. Martin Wolkewitz and his team from the Institute of Medical Biometry and Statistics at the University of Freiburg received part of the data. “We were able to discover that statistical errors had led to distorted results – caused for example by unrealistic model assumptions,” says Wolkewitz. “According to our results, Tamiflu does have a positive effect: anyone who is treated with it was able on average to leave the hospital sooner. However we were unable to verify an effect on mortality.” The authors of the study were however completely unaware of the errors: many researchers simply do not know about the nuances of statistical models, Wolkewitz explains. “Particularly in emotionally-charged issues such as medicines research we are able to help to provide information with the rational arguments of statistics.”

Banks want to evaluate the risk of investments as precisely as possible.
Photo: Deutsche Börse AG

Case study 2: Financial Market Research – how banks can improve assessment of risks

Banks have to deposit equity as a security, to be prepared for potential losses. The size of the security depends on the scale of the risk of the investment. “The aim of a bank is to estimate the appropriate amount of capital as well as possible – not too much and not too little,” financial mathematician Prof. Dr. Thorsten Schmidt from the Department of Stochastic Systems at the University of Freiburg explains. To do this, the common measure used is ‘Value at Risk’ (VaR) which is an amount that the potential losses of a bank are not expected to exceed with a probability for example of 95 or 99 per cent. The banks hereby attempt to evaluate future fluctuations in the value of their investments on the basis of historic data. “However, with the current processes, the actual rate of losses was higher than expected,” says Schmidt. Working with colleagues, the financial mathematician hopes to develop a solution: “We are working on refining the process so that banks can estimate the risks of their investments better – and we hope this will contribute to making the banking world slightly more secure.”

Nicolas Scherger


Anniversary celebration

The celebrations for the 30th anniversary of the seminar take place on Friday, 3rd November 2017 from 1.30 p.m. – 5.30 p.m. at the Institute of Mathematics, Eckerstraße 1, Room 404. Following speeches by Rector Prof. Dr. Hans-Jochen Schiewer and Prof. Dr. Josef Honerkamp, there will be lectures by Prof. Dr. Leo Held, biostatistician, from the University of Zürich, Switzerland, Prof. Dr. Rainer Dahlhaus, mathematician, from the University of Heidelberg, and Prof. Dr. Josef Teichmann, financial mathematician, from the Swiss Federal Institute of Technology (ETH), Zürich. Anyone who is interested is welcome to attend. Admission is free. No need to register.