Mock exams

Warning

The exams, their worked solutions and their LaTeX sources are assessment material and are deliberately kept out of the public repository, so they are not downloadable from this site. Instructors can request them from the author at info@profweisser-ai.de. This page documents what exists and how it is built.

These are practice papers, not the graded exam

The module (6 ECTS) is graded by one written exam at the end of the semester: 120 minutes, 100% of the mark. None of the eight papers below counts towards it — they exist so students can find out where they stand before that paper. The Final Mock Exam is deliberately built to its shape: same length, same structure, same weighting.

Eight practice papers are matched to the course rhythm: three full-length mock exams and five 60-minute short exams. Each is built from a single LaTeX source so the paper and its solutions can never diverge, and all numeric answers were verified programmatically.

Every paper exists in three formats:

  • Questions — the paper as students see it;

  • Solutions — the same paper with full worked answers, handed out together with the paper (so the discipline of working it closed-book first sits with the student);

  • Review deck — a Beamer deck for going through the exam in class, step by step.

The three full-length papers

Exam

Written after

Covers

Length

Mock Exam 1

after Ch 3

Ch 1–3

90 min · 90 pts

Mock Exam 2

after Ch 6

Ch 4–6 (+ light cumulative)

90 min · 90 pts

Final Mock Exam

after Ch 8

All chapters (weighted to Ch 7/8 and modules A9 and A7)

120 min · 120 pts

The final exam also exists in three parallel versions (A / B / C) — same structure and difficulty, different numbers — for seating variants or for a second attempt.

Marking

Each paper is built to be corrected quickly and consistently:

  • a marking table on the front page — per-problem maxima with a box for the awarded marks;

  • an “Answer key at a glance” page opening every solutions PDF: one row per sub-part with its expected result and its credit, so most scripts can be marked from that single page and the prose consulted only for partial credit;

  • a grading key after each problem, splitting the credit sub-part by sub-part and stating what earns partial marks and where error-carried-forward applies.

The answer-key page and the review deck’s “Marking at a glance” frame are both generated from one file per paper in keys/, so the two can never disagree.

make exam-check   # points add up, keys match, nothing leaks, variants differ
make exams        # rebuild the papers (runs exam-check first)

exam-check reads the LaTeX sources only. It fails if the sub-part points do not add up to the problem header or to the stated total, if the answer key misses a sub-part or awards different credit from the paper, if a solution or grading box sits outside the \withsolutions guard (which would print the answers on the student paper), or if two versions of the final exam share a computed answer.

The five 60-minute short exams

Alongside the three full-length papers there are five 60-minute exams (A–E), each three problems × 20 points, with the problems increasing in difficulty: P1 warm-up, P2 core, P3 advanced. They are the formative layer — short enough to sit in a spare hour, released one at a time as the material each needs is taught.

Short exam

Release after

P1 (warm-up)

P2 (core)

P3 (advanced)

A

after Ch 4

Learning problems, descriptive statistics (Ch 0 + 1–2)

Simple linear regression by hand (Ch 3)

Logistic regression, confusion matrix (Ch 4)

B

after Ch 5

Bias–variance, KNN regression (Ch 2)

Reading multiple regression output (Ch 3)

Cross-validation and the bootstrap (Ch 5)

C

after Ch 6

Conditional probability, Bayes, odds (Ch 0 + 0b)

Dummies and interactions (Ch 3)

Ridge and lasso (Ch 6)

D

after Ch 8

Honest model evaluation — leakage, splits (Ch 2 + 5)

KNN classification by hand (Ch 2 + 4)

Trees: Gini, splitting, pruning, forests (Ch 8)

E

after Ch 8

Reading describe() output, SE vs SD (Ch 0)

Polynomial regression, model choice (Ch 5 + 7)

Multiple testing: Bonferroni, Holm, BH (Ch 13 — now module A7)

The bold chapter is where each paper’s hardest problem sits, so the five are sequenced, not interchangeable: A needs Chapter 4, B Chapter 5, C Chapter 6, D Chapter 8, E module A7 (ISLP Ch 13). Because P1 and P2 reach back to earlier material, each paper is cumulative-to-date rather than a single-chapter test. They are also the only papers that touch the precourse material: A, C and E each open on Chapter 0 / 0b, while no problem in the three full-length papers cites either precourse deck.

Each ships as the paper, a solutions PDF carrying a grading key per problem (marks per sub-step, follow-through rules), and a review deck with a “common mistake” box per problem and a one-slide marking table. They are built by their own build.sh, not by make exams — see How they are built below.

Question design

  • Sub-parts are independent: a student who gets part (a) wrong can still earn full marks on (b) and (c).

  • Papers mix conceptual, mathematical and Python-interpretation questions — the latter show real output (a statsmodels summary, a confusion matrix, a CV curve) and ask what it means.

  • Each exam references the lecture slides it draws on, so revision can be targeted.

How they are built

Note

Mock_Exams/ is git-ignored, so it is absent from every clone of this repository — the commands below only work on a machine that already has the folder from the author. On a fresh clone there is nothing to cd into.

One source file produces both the paper and the solutions; the \withsolutions flag switches between them.

cd Mock_Exams/Mock_Exam_1_after_Chapter_03     # only if you have the folder
pdflatex -jobname=Mock_Exam_1 mock_exam_1.tex
pdflatex -jobname=Mock_Exam_1_Solutions "\def\withsolutions{1}\input{mock_exam_1.tex}"

The review decks are separate Beamer sources in the same folder (solutions_slides_*.tex) and compile with a plain pdflatex run — twice, for the navigation bar.

From the repository root, make exams builds the whole set on a machine that has the folder: 18 PDFs, namely

  • the three papers and the three final variants A / B / C, each as questions and, via \withsolutions, as solutions — 12 PDFs;

  • all six review decks, two passes each — 6 PDFs.

The target is incremental: a PDF is rebuilt only when its source is newer, and make -j is safe because every job writes under its own -jobname. On a machine without Mock_Exams/ — a fresh clone, for instance — it stops with a one-line message rather than a confusing LaTeX error.

The five 60-minute exams are the one exception. They live in Mock_Exams/Short_Mock_Exams_60min/ and are built by the build.sh that ships next to their sources (15 further PDFs: paper, solutions and review deck for each), deliberately kept out of the Makefile so that folder stays self-contained:

cd Mock_Exams/Short_Mock_Exams_60min
./build.sh

Where to go next