2026-09-12 –, Maschinenraum - C110 Language: English
In the light of recent high-profile cases of scientific fraud, being able to critically assess studies is increasingly important. Scientific publications, however, often act as polished interfaces, presenting a clean, coherent narrative, while hiding away the messy reality of the underlying raw data. Assessing a study, therefore, means looking past this interface and directly examining the data it is built on, which will be taught in this talk.
Science is the basis of decision- and policy-making, and it is generally a good idea to "trust the science." Recent high-profile cases of scientific fraud, however, have demonstrated that scientists are only human too, and sometimes, for various reasons, individual scientists falsify their research. Accordingly, we all should be able to scrutinise and independently assess scientific studies, spotting potential cases of manipulation. Scientific publications, however, often act as polished interfaces, presenting a clean, coherent narrative, while hiding away the messy reality of the underlying raw data. Assessing a study, therefore, means looking past this interface and directly examining the data it is built on.
In this talk, we will, therefore, focus on the raw data of scientific research and their forensic examination. To get started, we begin with a brief introduction to the structure of a typical scientific paper and the process of modern scientific publishing. Afterwards, we will take a look at a number of recent high-profile cases of scientific fraud together in a hands-on manner. We will thereby establish practical guidance and concrete statistical tools for identifying potential cases of data manipulation which may warrant further examination. These approaches will enable and empower audience members to independently perform forensic plausibility checks on scientific data, as well as provide a starting point for their further, independent study of additionally provided resources. Since no tool or technique is perfect, we will also talk about the limitations of the presented approaches, as well as ethical considerations when performing such analyses.
This talk is directed at everyone with an interest in scientific research and everyone who enjoys critically assessing datasets for plausibility. Prior knowledge in statistics, data science and data visualisation are certainly advantageous, but not required, as all necessary theoretical foundations will be introduced during the talk.
PhD graduate in biomathematics, now working in industry, interested in data analytics, (frequentist and Bayesian) statistics, functional programming, verifiable code, and lots of other obscure nerd stuff.