Broader course for all
Limited to 20 participants
Roger Wolf (KIT)
Abstract
The mathematical discipline of statistics can be very abstract. This may impair the understanding of how decisions are reached and bares the vulnerability to conscious or unconscious misuse. This course is meant for KCETA students, which are confronted with the need for statistically sound analyses of scientific data or are just interested in a more intuitive understanding of statistical concepts without counting this to their daily work. I will discuss a selection of basics that are most important for statistical data analysis in the area of physics or natural science in general. Since I assume that the bare information is already known, this will not be a usual introductory course on the topic. I will rather pick topics and discuss or question them to arrive at a deeper understanding of their actual meaning. A loose collection of topics may comprise the following (classical) points: Bayesian vs. frequentist probability interpretation, mathematical modeling, the likelihood concept, parameter estimation, hypothesis testing, and limit setting on physics parameters within models. A question that I consider extremely intriguing is: "What is a physics measurement and under what conditions does a parameter estimate within a model turn into a model-free truth?". I will base the discussion on hands-on examples formulated in Jupyter-notebook, i.e., access to a Jupyter-server with the following software installations will be required to make maximal use of the course: python3, numpy, pands, matplotlib, kafe2. Those interested in the course, who don't know how to fulfill these requirements may logon to a dedicated profile on the bwJupyter Hub that will be provided before the beginning of the course.