🧐 Comments of the W04 Summative and model solutions

When just being correct is not enough

Author

This page contains an analysis of your submissions for the ✏️ W04 Summative, worth 10% of your final grade.

πŸ“Š Submission statistics

Total Accepted
Assignment
Not pushed
to GitHub
(empty repo)
% of
enrolled students
who submitted
61 58 4 88.52%

I hope the seven of you who didn’t submit were preparing yourselves for the next summative! 🀞

Score distribution

🚧 WIP 🚧

  • We have not yet finished marking all submissions
  • I will update this part of the page later

πŸ“ Solution

The majority of submissions I inspected managed to produce a valid predictive model and assessed the MAE on the training and test sets! πŸŽ‰

But note that our marking criteria emphasised the notebook’s organisation and the code’s clarity. A valid, correct solution would only get you a 50% score. This is because we want to encourage you to develop good habits early on in the course.

Model solutions

Some solutions stood out for their clarity and organisation and reached 100/100! I provide them her with permission from the authors. I hope you can learn from them!

Submission by Source
Notebook
Rendered
page
Amara submission1.qmd submission1.html
Tania submission2.qmd submission2.html
Anonymous submission3.qmd submission3.html

πŸ“Œ Common mistakes

We haven’t finished marking all submissions, so there might be more things that our markers will spot that I haven’t spotted. But overall, most submissions were good! You used markdown well, had a nicely rendered HTML report, and documented your thought process well. By the way, you didn’t even have to write a lot. If you managed to be concise yet convey your thought process, that’s great!

Therefore, the list of the most common mistakes relates mostly to the organisation of the notebook and to the failure to follow the instructions in the assignment:

CM1: No HTML file

As per the instructions, you have to provide the source, .qmd file, as well as the rendered .html file. The HTML file is the one we use to mark your assignment, as it makes it easier for us to navigate through your thoughts and your code.

Another reason we enforce the rendering to HTML is that it forces you to ensure your notebook is reproducible. If you can’t render it to HTML, it means that you have some missing packages or that you placed the code in the wrong order.

Markers were instructed to:

  • If the resulting HTML file is not self-contained OR there is no HTML file, deduce 15 marks from Criterion 01! Harsh!

CM2: Underwhelming use of markdown

As explored in the 🧐 W03 Formative Analysis, it is important to produce a nice report. We expect you to use good markdown practices: good headers, documentation of your thought process, etc. If you fail to create a nice markdown, you will lose marks.

Markers were instructed to:

  • If the submission had headers that were confusing or didn’t render properly in HTML, deduce 5 marks from Criterion 01.
  • If a student had a lot of output printed out (forgot to suppress warnings when loading libraries), deduce 5 marks from Criterion 01.

CM3: Failure to guide us through your thought process

Markers were instructed to:

  • If there’s barely any text, it’s mostly just code chunks and a few sad headers; deduce 20 marks from Criterion 03! (this is important!)
  • In the final steps, if the student made vague comments such as β€˜the model didn’t fit well’ but did not explicitly express why they think so, nor put the residuals into perspective properly, deduce 10 marks from Criterion 03.
  • If there was no conclusion, the text just restates what the numbers say ( β€˜the MAE in training is lower than testing’ and that’s it), this is worse than the case above. Therefore, deduce 15 marks.