Datasets
All DataLab datasets are synthetic — generated by seeded, reproducible scripts — but deliberately shaped like real survey and education data, missing values and all. They sit inside EURIDICE’s sector focus: Communication & Media and Education & Culture.
They are preloaded in every notebook under data/, and downloadable here
for use in your own tools (Excel, R, anything).
EU Media Habits 2026
Section titled “EU Media Habits 2026”Survey of 624 respondents in 8 European countries on media and news consumption. Used in Track 2 — Exploratory Data Analysis.
⬇ Download eu_media_habits.csv · in the lab: data/eu_media_habits.csv
| Column | Meaning |
|---|---|
respondent_id | Unique respondent number |
country | Respondent’s country (8 countries incl. Iceland, Portugal, Ukraine) |
age, gender | Demographics (16–79) |
household_income_band | Income bracket — sensitive; often refused |
hours_tv | Daily TV hours — missing at random (~4%) |
hours_social | Daily social media hours |
hours_news | Daily news consumption hours |
main_news_source | TV, social media, online news, radio or print |
trust_news | Trust in news media, 1–10 — missing not at random (~14%) |
pays_for_news | Pays for a news subscription — co-refused with income |
The missing values are the point
Section titled “The missing values are the point”This dataset carries three deliberately different kinds of missingness, so learners can tell them apart:
hours_tv— MCAR: random dropout with no pattern, the harmless kind.trust_news— MNAR: people who distrust the news refuse to rate their trust, so the answer itself causes the gap. The survey under-reports deep distrust by roughly a third.household_income_band+pays_for_news— correlated refusal (≈0.75): the same privacy-guarded respondents skip both money questions.
Running dropna() on this file deletes 32% of respondents — and not a
random 32%.
Answer key: eu_media_habits_answer_key.csv holds every respondent’s
true trust rating, including those who refused. A real survey can never have
this; because the data is synthetic, Track 1’s missing-data lesson uses it to
show exactly how much the refusals distort the finding.
⬇ Download the answer key
Student Digital Skills
Section titled “Student Digital Skills”900 learners with study background, learning habits and device access; target column records whether they passed a digital-skills certification. Used in Track 3 — Machine Learning Fundamentals.
⬇ Download student_digital_skills.csv · in the lab: data/student_digital_skills.csv
| Column | Meaning |
|---|---|
student_id | Unique learner number |
country, age, study_field | Demographics; fields from ICT to Humanities |
hours_online_learning | Weekly hours of online study |
digital_skill_score | Skills assessment score, 20–100 |
has_own_laptop | 1 = yes, 0 = no |
prior_ict_courses | Number of previous ICT courses (0–5) |
passed_certification | Target: 1 = passed, 0 = failed |
Licence & provenance
Section titled “Licence & provenance”Generated by IIIM for the EURIDICE project
(scripts/make_datasets.py in the DataLab repository — fully reproducible).
Licensed CC BY 4.0: use them freely in your own teaching, cite
EURIDICE DataLab / IIIM.
Co-funded by the European Union under the Digital Europe Programme — projectEURIDICE, Grant Agreement No. 101123121. Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the granting authority can be held responsible for them.
Built by IIIM — Icelandic Institute for Intelligent Machines. Learning materials are licensedCC BY 4.0.