Project

The practical part of the exam is a data analysis project on an assigned dataset, carried out in two parts and presented at the exam. The work happens in a GitHub repository forked from the course template; if you would rather not use GitHub, a zip by e-mail is accepted.

Template repository Parte 1 Parte 2

How it works

  • Groups of one to three students.
  • Register the group with the course form (MS Forms, linked on Teams) within the first two weeks of the course. The form is how the groups are formed; it is not the place where the dataset is chosen.
  • Each group is then assigned a dataset by e-mail, in the third week. The analysis is yours to design: the template says what a complete analysis contains, and you write the questions.
  • You fork the template and work in your fork. Three ways to hand a part in, whichever suits you:
    • fork public and send the link;
    • fork private and add antoninofurnari as a collaborator;
    • or send a zip of the repository by e-mail.
  • The presentation at the exam doubles as the oral examination: every member speaks and answers questions about the analysis.
Released Due
Part 1 — exploratory and inferential analysis third week (tentative), with the dataset reviewed in class at the first written test
Part 2 — explaining, predicting and representing session 21, before the break end of January
Presentation — early February, before the first exam session

Part 1 is looked at in class during the session of the first written test: the teacher and the tutor go round the desks and read what each group has. It does not have to be finished — the point is to catch a wrong turn while there is still time to correct it.

Part 2 is walked through in class before the winter break, so you can work on it while the last lectures are still running. It is not tied to the second written test, which is a test only.

Ground rules

  • The notebook must run top to bottom in a clean environment. If it does not, the submission is returned before it is marked.
  • Data files are never committed. The repository explains how to obtain them.
  • Strip notebook outputs, or keep the notebook under 5 MB.
  • A contribution breakdown and a declaration of how generative AI was used are both mandatory.
  • A submission consisting only of slides is not marked.
  • A negative result, argued with evidence, is worth full marks. An analysis that honestly concludes “the data cannot answer this” is a good analysis.