Source and coverage notes
Scope: the nine current Canvas lecture pages and their linked presentations, plus automatic transcripts of the lecture recordings (2–9), the 12 papers on the Canvas literature list and two official example exams. Canvas lecture pages/file metadata checked 11 October 2026 at 11:40 CEST (Europe/Amsterdam).
What the site covers
Detailed notes cover the teaching content in the presentations, including demonstrations and guest-lecture cases. Progressive/repeated slides are consolidated. A few original slide figures are shown where they help; other figures and interactive datasets are explicitly illustrative.
Transcripts for 2–4 help explain demonstrations. Update 11 October 2026: the Canvas recordings of lectures 5–9 were transcribed automatically and compared with the slides. Each lecture page now has a “From the recording” section listing examinable points that are not (fully) on the slides, and several notes were corrected. Lecture 1 has no recording on Canvas, so its notes remain slide-based. Automatic transcripts can mishear terms; check the recording for exact wording. A course literature page summarises the 12 papers on the Canvas literature list, and a question bank adds exam-style questions for every lecture.
Presentations
| Lecture | Presentation used | Canvas file ID | Pages |
|---|---|---|---|
| 1 | lecture1.pdf | 10779930 | 69 |
| 2 | lecture2.pdf | 10628872 | 59 |
| 3 | lecture3.pdf | 10628777 | 65 |
| 4 | lecture4.pdf | 10628856 | 73 |
| 5 | lecture5_update.pdf | 10628829 | 60 |
| 6 | lecture8.pdf (older numbering) | 10628887 | 97 |
| 7 | lecture7.pdf | 10628764 | 24 |
| 8 | BigDataHealth2025.pdf | 10628765 | 76 |
| 9 | 11. user generated data.pdf | 10628779 | 137 |
Lecture 6’s Canvas page explains that its linked deck used to be lecture 8. An older renumbered version of the deck (lecture6.pdf in the Canvas course files) is used as supplementary concept slides because it develops Gini/tree thresholds and numerical PCA in more detail. The two complete decks overlap; they are not two extra lectures.
Lecture 1 uses the current Canvas-linked lecture1.pdf. Slide-based administrative information is identified as such; check Canvas for current deadlines.
How the example exams were used
- Practice_exam_data_science.pdf: internally dated 2024; 30 MC and eight open questions. It emphasizes definitions, method choices, feature operations, output/plot interpretation and brief justifications.
- test-exam-DSH-2022_2023.pdf: 30 MC and seven open questions, including short R tasks. This is why the site includes code tracing and a small function/pipeline rather than conceptual questions alone.
- How to study for the exam.pdf: emphasizes understanding key ideas, methods, formulas, assumptions and examples rather than memorizing every detail. It says most questions concern lectures and a smaller number concern course literature; the practice exam itself covers the slides. The course literature page covers each paper at that level: its question, approach and main findings.
The new practice problems use new situations and counts. They do not reproduce the official exam prompts/figures. A limited presence in an old sample is not evidence a current lecture topic can be skipped, particularly for lectures 8–9. Some older exam study cases differ from the current slide decks.
Corrections and careful interpretations
The notes explicitly clarify z-scoring versus normality, min–max versus log transformation, probability versus odds, ratio=1 as the null comparison, likelihood versus R², training versus test overfitting, and hazard ratio versus absolute probability. Rule-of-thumb model comparisons are presented as hypotheses to validate, not guarantees.
Mathematical expressions are written as readable text and R code symbols are preserved. Answers stay collapsed until you open them.
Verification
Interactive pages use only embedded JavaScript and CSS, with no external libraries. Progress stays in your browser unless you turn on device sync, which stores it on this site under your sync code; the host (Cloudflare) counts page views. The calculations were checked, the R code in the questions was run, links were checked across the site, and pages were reviewed at phone and desktop widths.