Conferences

Many lecturers use surveys in order to teach elementary statistics or survey methodology – due to various reasons, there are often typical limitations associated with these efforts: No use of “real” data but rather a multitude of pre-cleaned samples to fit certain topics, uncoordinated use of tools/software, introduction to the actual sampling process late in the curriculum. In our experience, this has often led to students being unable to properly conduct surveys or use real data from official surveys when faced with this task.

The aim of this talk is to give an overview of the redesigned statistical education at the Freie Universität Berlin with the particular aim to overcome the above limitations: The program consists of 2 main blocks: Elementary statistics with extensive use of a software called “statistical lab” (see also: (www.statistical-lab.com) that was specifically designed to support ease of use but also professionalism through being based on the language R. The 1-year elementary-curriculum also heavily builds on survey data collected during the course thus exposing students to the challenges of survey statistics early on. At the advanced level the students naturally transition from the “statistical lab” software to R and SAS and from simpler to more complex surveys including the use of the full but anonymized German Microcensus provided via a Campus agreement.

A key part of the whole program is that almost all exams (including those for elementary statistics with over 300 students per class) are based on the use of statistical software and electronic submission of the results during timed tests. In addition the whole course could be followed over the internet. Both has been a particular administrative challenge but also a key driver for the success of this program and we will give details on the tools and processes used to make this possible.

Given that the first cohort of students has fully completed this program we would also like to reflect on typical pitfalls and challenges encountered as well as our key insights that might be of use for other statistical programs.